April 2017 ISSUE 56
Microgrids the future of Electrical Energy Grid page 08
Proportional and Simultaneous Myoelectric Control of a Robotic Arm
Sensorless Position Control of a PMSM for Steer-by-Wire Applications page 32
page 24
EMG Signal Analysis and Finger Movement Classification
page 16
main sponsors
April 2017 ISSUE 56
Contents 03 04 From the Editor
From the President
24
32
Proportional and Simultaneous Myoelectric Control of a Robotic Arm
www.coe.org.mt
08 Microgrids the future of Electrical Energy Grid
16
Cover Image
EMG Signal Analysis and Finger Movement Classification
Sensorless Position Control of a PMSM for Steer-by-Wire Applications
EMG Signal Analysis and Finger Movement Classification Page 16
Editor
Mr Stephane Role
Editorial Board
Inġ. Norman Zammit Eur. Ing. Inġġ. Pierre Ciantar Prof. Dr Inġ. Robert Ghirlando
Chamber of Engineers, Professional Centre, Sliema Road,Gzira, GZR 1633, Malta
Email: info@coe.org.mt Web: www.coe.org.mt
© Chamber of Engineers 2017. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopy, recording or otherwise, without the prior permission of the Chamber of Engineers - Malta. Opinions expressed in Engineering Today are not necessarily those of the Chamber of Engineers - Malta. All care has been taken to ensure truth and accuracy, but the Editorial Board cannot be held responsible for errors or omissions in the articles, pictographs or illustrations.
Design by: Printing: Print It Ltd. |
Distribution: Maltapost Plc.
April 2017 ISSUE 56
1
| TC11-02E |
Machine control for Industrie 4.0 with TwinCAT
www.beckhoff.com/Industrie40 Beckhoff provides the ideal foundational technologies for Industrie 4.0 and Internet of Things (IoT) applications via standard PC-based control. With the TwinCAT engineering and control software, machine control systems can be extended to support big data applications, cloud communication, predictive maintenance, as well as comprehensive analytical functions to increase production efficiency. As a system-integrated solution, TwinCAT IoT supports standardised protocols for cloud communication and enables the easy integration of cloud services right from the machine engineering stage. In addition to fault analysis and predictive maintenance, TwinCAT Analytics offers numerous opportunities to optimise machines and systems in terms of energy consumption and process sequences.
Ivy, Flat 1, Salvu Barbara Street Mosta, MST2433, Malta Phone: +356 21420655 info@jmartans.com www.jmartans.com
From the Editor Dear Readers, I welcome you all to the 56th issue of Engineering Today. In this issue we will be taking a look at two student final year projects and research done by engineers from the Department of industrial Electrical Power Conversion at the University of Malta. One of the student final year projects is a winner of the Chamber of Engineers’ Engineering Student Project Awards held last December while the other student is a winner of this year’s MGPEI MCAST Final year project. Coincidentally, these projects are biomedical engineering projects and both of them show the advancement of this field in Malta. Prof. Dr Ing. Cyril Spiteri Staines explains the future of the electrical energy grid in the first article of this issue. In this interesting piece the concept of microgrids are introduced. Microgrids house a large number of benefits over the conventional grid system. With an increasing dependency on a distributed electricity generation scenario, microgrids can increase efficiency and better manage the power distribution system.
Per usual, I would like to thank all the contributors to this issue for sharing their expertise and experience with the magazine and its readers. I would like to take the opportunity to congratulate all the victors of the final year project awards. Any articles for our next issue can be submitted on the Chamber of Engineers’ website http://www.coe.org.mt/ publications. As a final note, I would like to express that I am always open to feedback to better this publication. While I greatly appreciate the quality and submissions from engineers in academia, I would also like to appeal to engineers in other sectors to share their knowledge and experiences with the reader base. From the next issue “Engineering Today” will be heading in a new direction to better showcase our achievements as engineers and the good practices we develop to meet our daily challenges.
Mr Adrian Von Brockdorff shows us how he uses electrical signals generated by the actuation of our hand muscles to determine which finger is being moved. Mr Brockdorff built a mechanised hand and used it to display the output. Similarily, using electrical signals from muscular activity, Mr Christian Grech has developed a model to convert these signals into shoulder angles to control a robotic arm in real time. Our final article was submitted by Mr Kris Scicluna. The project he describes makes part of his PhD studies carried out with Department of industrial Electrical Power Conversion at the University of Malta. Mr Scicluna has developed a model to simulate a sensorless speed or position estimation for a permanent magnet synchronous motor, to be used in a steerby-wire scenario. Reducing the number of sensors should increase the reliability and robustness of the system.
Mr Stephane Role B.Eng.(Hons)(Melit), M.Sc Bio.Eng.(Imp Lond) The Editor, Engineering Today, Chamber of Engineers
April 2017 ISSUE 56
3
From the President Dear Colleagues, This edition of our publication is being issued after the Annual General Meeting that was held on the 24th of February 2017 as required by the statute. This year the number of nominations for Council required the holding of an election. 1. Points from the Presidential Address during the AGM: The Council met for eighteen times during the previous administrative year (Feb ’16 – Feb ’17) and it was intensive work on various issues throughout the whole year. Council support: The Council cannot work in isolation and therefore it was deemed appropriate to recognise the important work that is carried out by those fellow engineers who devote their time towards the Chamber and support the Council.
•
Ing Helga Pizzuto - Vice-President of the Federation of Professional associations representing the Chamber together with Ing. Saviour Baldacchino.
•
Prof Dr. Ing. Paul Micallef as the Chairperson of the CPD subcommittee and its members, Ing. Joe Camilleri (Chairman – MGPEI), Dr. Ing. Daniel Micallef from the Chamber Council -. Ing Helga Pizzuto shall be consulted due to her involvement in preparing the framework for the Federation.
•
Ing. Johan Psaila our Treasurer on the BRO.
•
Prof Robert Ghirlando, Dr Ing. Victor Buttigieg, Ing. Pierre Ciantar members within the Ethics and Disciplinary committee under the chairmanship of Ing. Alex Galea.
•
Mr Stephane Role as editor for Engineering Today, who together with Prof Robert Ghirlando, Ing Pierre Ciantar and myself form the Editorial Board of our flagship publication.
•
Dr Ing. Daniel Micallef for his contribution towards the introduction of national standard for Green Roofs
•
Marcelle Abela – Administrative Secretary
•
Marthese Attard – Training Executive
•
Alberto Cachia – Communications Officer
•
Internal Auditors
•
Panel of Judges for the MEEA.
4
Issues tackled during 2016 During 2016, the Council addressed several issues and these included:
•
Amendments to the Engineering Profession Act Chap 321 – A task force from the Council was engaged in a collaborative exercise with the Ministry for Transport and Infrastructure to amend the Engineering Act in line with the requirements of the Mutual Recognition of Qualifications Directive 2005/36/EC and The Services Directive 2006/123/EC.
•
Review of the Engineering degrees from the UoM and MCAST - The Chamber of Engineers had been following up with the Engineering Board on the issue of engineering degrees from MCAST vis-à-vis the Engineering Warrant. The Engineering Board appointed international reviewers to carry out a review of all the Institutions that provide for engineering degrees in Malta and the conclusion of this process will set up a benchmark that needs to be attained by any Institution to have its degree recognized by the Board and hence allowing its graduates to apply successfully for a warrant. The process has now been concluded and a report with recommendations has been forwarded to the Hon. Minister of Transport and Infrastructure for the recommended actions. The relevant parts of the report were also issued to University of Malta and MCAST. The Chamber of Engineers is following up the conclusions with a meeting that has been set up in May with the Hon. Minister and the Engineering Board.
•
New Legislation Regarding the Safe Use of Work Equipment – Legal Notice L.N. 293 of 2016 – OHSA – In L.N. 293 of 2016, the term competent person is given an ambiguous definition which could be interpreted as not meaning an engineer, architect or medical practitioner. OHSA officials are advising owners of lifting equipment that they do not need to produce a certificate from a warranted engineer, but from a competent person, whatever that may mean. Employers were also advised that the competent person could be the owner himself or one of his employees; this goes against all sense of ethics and worker safety. Following an exchange of correspondence and a meeting with the OHSA Board, the CoE has made its position clear with the OHSA in the sense that:
•
Once the OHSA decides to ask the employer to provide a certificate at the employer’s expense, such certificate
must be issued and signed by an Engineer, Perit or Medical Practitioner.
Minister” [Engineering Profession Act Chap 321]. International Sphere
•
The OHSA Board was referred to LN 354 of 2012 Practice of Engineering (Definition) Regulations, 2012 where this subsidiary legislation defines the practice of engineering that includes certification of engineering systems and processes.
•
The issue of conflict of interest, where owners are being allowed to certify their own equipment was also raised. This is also an important issue when it comes to worker safety and we stressed that rather than a person from the owner’s organisation being allowed to carry out this task, the requirement should be that the appropriate certification is issued by an independent engineer.
Issues of the Kitchens without external ventilation PA and BRO involvement in kitchens without external ventilation. These are allowed by the PA subject to particular canopy design. CoE brought up serious concerns on this issue with the BRO who are willing to proceed with issuing guidance notes regarding the design of kitchen extractor hoods in order to eliminate ‘ventless hoods’ concept. Commencement Notice document As part of the CoE’s input to the updating of the commencement notice procedures, feedback received from members was compiled in a document and forwarded to Hon. Deborah Schembri. The issue shall be followed up by the new Council. Continuous Professional Development As per Act 321 – CPD shall mean a structured and organized activity, planned and carried out by warrant holders so as to maintain and update their knowledge of the art and science of their profession, and enhance their ability to exercise their profession, in the context of an ever-expanding body of professional expertise, as well as to keep abreast of regulatory and ethical developments and to motivate professionals to engage in lifelong learning relevant to the safe and effective practice of their profession. The CPD committee shall work towards finalising the framework for implementation by end of 2017. Code of Ethics A new edition of the Code of Ethics has been finalised and shall be shared amongst the members by Q2 of 2017. Although the Code of Ethics as issued by the CoE is entrenched in the law, as one of the amendments to the law, there shall be more specific references to it. “20. (2) Warrant holders are required to follow the Code of Ethics issued by the Chamber of Engineers and approved by the
•
Common Framework Principles – initiated by the ECEC and supported by FEANI. Two meetings were held in Vienna in this regard and a task group consisting of Robert Ghirlando, Anthony Cachia and Daniel Micallef, worked on the proposal document stressing the fundamental principle that a degree in Malta that qualifies for a warrant is to be of 240 ECTS credits irrespective of whether it is a 3, 4 or 5 year level 6 degree.
•
FEANI Annual General Assembly in Stockholm – Sweden – In 2016 the general assembly continued its focus on the common training framework being proposed by FEANI which complements the ECEC initiative. Malta kept the same stance as in other instances where it was made clear that there is no mature candidate route to becoming an engineer and hence this should be reflected in the document for common training framework principles.
Renewal of Corporate/Collaboration agreement with BOV BOV renewed its corporate sponsorship agreement for another year with the future introduction of a monetary award to the finalist of the UoM final engineering students’ projects winners. 2. Executive Council Composition for 2017-2018: Following the Council meeting of the 28th February, the following is the new Council composition for the year Feb 2017 - Feb 2018: President:
Ing. Norman Zammit
Vice President:
Ing. Saviour Baldacchino
General Secretary:
Dr Ing. Daniel Micallef
Treasurer:
Ing. Johan Psaila
PRO:
Ing. Jason Vella
Activities Secretary:
Ing. Mike D'Amato
Membership Secretary:
Ing Alex Galea (also chairperson of the ethics committee)
Assistant Membership Secretary:
Ing Anthony Cachia
Secretary for International Affairs:
Prof. Robert Ghirlando
Member:
Mr Stephane Role (also confirmed as Editor and assistant PRO)
Student Member:
Mr Alex Tanti
April 2017 ISSUE 56
5
ABSOLUTE SPILL SOLUTIONS
WHATEVER YOU DROP
OILS & LUBRICANTS
FUELS
CHEMICALS & EMULSIONS
BLOOD & OTHERS
Oilex is a non-toxic universal organic binding agent that is able to quickly and effectively absorb all kinds of oils, fuels, chemicals, emulsions, blood and other polutants. THE ABSORPTION RATE OF OILEX IS 100% • Absorption capacity is up to 3-times of its own weight
• 100% organic fibre without artificial additives
• Can be re-used until it’s completely saturated
• Antitoxic for humans, animals and the environment
• With only app. 134 grams/liter, Oilex is extremely light and
• No contamination of soil and groundwater
easy to handle
• Cost-effective disposal after application due to its low weight
• Effectively absorbs substances and doesn’t release them under increased pressure
• No long-term physical of visual effect on the environment • Hydrophobic and long-term floatability
PRO TECH Engineering solutions Ltd 191, Arcade Street, Paola, PLA1211 | Tel. 2180 1129 | info@protechmalta.com
www.protechmalta.com
From the President Continued
3. Who is an Engineer in Malta? Since 1987, the Engineering Profession in Malta has been regulated under Chap 321 of the Laws of Malta ENGINEERING PROFESSION ACT. The principal act is followed by a number of legal notices (subsidiary legislation) that regulate further the work that falls within the responsibilities of an Engineer. Chap 321 clearly states that for a person to be referred to as an 'Inginier', s/he has to be in the possession of a warrant issued by the warranting board which, under the principal act, is the Engineering Board. The warranting board may also issue a temporary license in lieu of a warrant for those persons that qualify under the act and performing engineering services in the Maltese Islands for a short period of time. The Principal Act also indicates that "Any person who, not being the holder of a warrant, assumes or uses the designation "Inginier" or its abbreviation "Ing.", or in any manner indicates that he is entitled to exercise the profession of engineer shall be guilty of an offence and shall, on conviction, be liable to a fine (multa) not exceeding two hundred and thirty-two euro and ninety-four cents (€232.94) and in respect of a second or subsequent offence to imprisonment for a term not exceeding three months or to both such fine and imprisonment". Furthermore, the Act also qualifies further that, “Any person who, not being the holder of a warrant or a special licence or who is otherwise not entitled to practice temporarily in accordance with the provisions of this Act, practices the profession of engineer shall be guilty of an offence and shall, on conviction, be liable to a fine (multa) of not less than four hundred and sixty-five euro and eighty-seven cents (€465.87) but not exceeding nine hundred and thirty-one euro and seventy-five cents (€931.75), and in respect of a second or subsequent offence to a fine (multa) of not less than six hundred and ninety-eight euro and eighty-one cents (€698.81) but not exceeding one thousand and one hundred and sixtyfour euro and sixty-nine cents (€1,164.69) or to imprisonment for a term not exceeding six months or to both such fine and imprisonment."
Hence, as a Chamber of Engineers, our appeal goes to those individuals and organisations that are overlooking this fundamental principle either due to ignorance of the provisions within the Engineering Act or maliciously posing as Engineers even through just the presentation of a business card. The Principal Act and its subsidiary legislation indicate therefore that the practice of engineer is a regulated profession which embodies scientific and technological principles, in view of the over-riding need to protect public interest, particularly in relation to issues of public health and safety, protection of the environment and, protection of cultural heritage, and arising from the design, and supervision of engineering works, and therefore various Laws of Malta reserve relative tasks to be undertaken only by a holder of a Warrant. The practice of engineering consists in the activities of design, specification, development, installation, commissioning, certification, operation, maintenance and decommissioning of any mechanical, electrical, chemical process, information and related systems, and such activities shall be carried out by or under the authorisation and guidance of an engineer who is a warrant holder or a partnership, competent in the field as recognised by the Board. As stated earlier, our appeal goes to those individual, governmental and private organisations that are not abiding by the requirements of the legislation and that are employing individuals who are not engineers in engineering positions and giving them the title of Inginier or Engineer. Also, the attention is drawn to those individuals who present themselves in this manner. Irrespective of whether you have an engineering degree or not, the only legal instrument that defines the Engineer is the Engineering Warrant.
Yours Sincerely,
Inġ. Norman Zammit B. Elec. Eng. (Hons.), M.Sc. (Brunel), Eur. Ing., CBIFM President, Chamber of Engineers
April 2017 ISSUE 56
7
Microgrids – the future of Electrical Energy Grid Cyril Spiteri Staines, Alexander Micallef, Maurice Apap and Daniel Zammit Department of Industrial Electrical Power Conversion, Faculty of Engineering, University of Malta cyril. spiteri-staines@um.edu.mt
ABSTRACT Over the past years, there has been a shift from centralised power generation to distributed generation. Advancements in distributed energy generation, storage and energy management has led to the development of two new concepts for electrical energy networks, namely smart grids and microgrids. A smart grid monitors and optimally manages the energy flow in a network. A microgrid is a sub-power system formed from various types of generation sources (both conventional and renewable) and storage systems which work together to allow stable and efficient energy supply even when disconnected from the main electricity network. Microgrids shall be an essential part of the future electrical generation and distribution systems. Keywords: Electrical Grid, Microgrids, Smart-grids, AC Microgrids, DC Microgrids, Hybrid Microgrids, Energy Storage.
8
1 INTRODUCTION Microgrids are a relatively new concept in power engineering and are considered as the basic building block for the future electrical networks. Over the past 10-15 years, there has been a shift from centralised power generation (Figure 1) to power generation from distributed energy sources (Figure 2). This has taken place due to the increase in number of renewable energy sources (RES) at domestic, commercial and industrial sites. This increase in RES penetration is expected to be more significant in the coming years. The RES can includelarge solar/wind farms or even smaller systems such as residential PV panels installed on rooftops. The evolution from a centralized to a distributedpower system configuration is taking place on a worldwide level. This article shall attempt to give an insight on the power system network configurations of the future. Advancements in distributed energy generation, storage and energy management has led to the development of two new concepts for electrical energy networks, namely smart grids and microgrids [1],[2],[3]. Through a dedicated ICT infrastructure, a smart grid monitors and optimally manages the bidirectional energy/power flow in a network from the different forms of generation sources (conventional/ renewable) to meet the power demand from the loads. A microgrid is a sub-power system consisting of a cluster of small generation sources (both conventional and renewable) and storage systems working together to allow stable and efficient operation to supply energy to the cluster’s loads even when disconnected from the main electricity network. (An example of this is a group of households cooperating together to form a small stable electricity network between
Figure 1:
Traditional Power System withUnidirectional Power Flow from Power Station outwards.
each other under the supervision of the network distribution operator). Therefore,microgrids can be integratedwith present electrical network architectures but can also shape the future smartpower systems. Figure 2 illustrates the smart grid concept and shows how microgrids fit into this futuristic scenario. The ICT infrastructure of the smart grid is a key enabler to achieve connectivity for all devices in the electrical network. This will allow for monitoring and storage of information which the power system can use for decision making in order to increase energy efficiency, make the grid more reliable, implement peak load shaving, trigger protection systems during faults and also communicate with microgrids present in the electrical network. 2 WHY MICROGRIDS? The increasing demand for clean, reliable and affordable electrical energy is changing the existing electricity generation and distribution scenario. One of the major aims of the microgrid is to combine the benefits of non-conventional/ renewable, low carbon generation technologies and high efficiency generation systems (e.g. CHPs). Microgrids can: enhance local supply reliability; reduce feeder loss; offer local voltage support; increase efficiency; correct for voltage sags and offer uninterruptible power supply. 3 TYPES OF MICROGRIDS Microgrids can consist of two main architectures: an AC microgrid or a DC microgrid [4]. Due to the widespread use of AC electricity and the off-the-shelf availability of AC equipment, AC microgrids do not have major infrastructural challenges hindering their deployment. On the other hand, DC microgridsare attractive since they offer advantages over their AC counterpart suchas: lower conversion losses due to
Figure 2:
Smart Grid Concept with Distributed Energy Sources, Energy Storage and Microgrids.
April 2017 ISSUE 56
9
Welcome the new era in
WATER HEATING
Modular system that can cater for any demand ranging from small to large residences, under-floor heating and pool heating Ideal for hotel industry and process hot water in factories Can be coupled to any heat source like oil, gas, pellet boiler, heat pump etc. Latest technology with unsurpassed efficiency levels Unbeatable performance when coupled with solar collectors Hot water on-demand! Legionella-free
WÜRTH Centre, Mdina Road, Zebbug, Malta Contact us on 21494604 or email us on solar@wuerth.com.mt
Microgrids – the future of Electrical Energy Grid Continued
Figure 3:
DC Powered Home Concept.
less conversion stages; grid synchronisation is not required; and they don’t experience voltage phase or frequency issues or other power quality issues that occur on AC grids (e.g. harmonics). Currently research is being carried out to develop the required technology to integrate DC systems in the present electricity networks. DC microgrids are deemed to be the future type of electrical grids [1] for building services. This is in light of the fact thatmost electrical loads nowadays actually use a DC source of power to operate (e.g. Laptops, smart phones and most consumer electronics; Inverter driven loads such as A/Cs, washing machines, fridge/freezers; LED Lighting; and Electric vehicles.) Figure 3s hows a home which makes use of a DC bus for supplying its loads and interconnection with RES and storage systems. The advantages offered by DC microgrids shall also help towards the realisation of nearly-zero energy buildings. In addition, the DC microgrid concept is not limited to individual buildings but it can also be extended to form a larger DC power distribution network linking a number of buildings. 4 DC MICROGRIDS The two voltage levels currently being considered for DC microgrids applied to building services are 24V and 380V [4],[5]. The 24V DC bus can be used for small scale (and low power) applications which are mainlyfound in residential buildings. For larger power levels,the 380V DC bus voltage is considered as the most appropriate voltage level.The higher bus voltage is therefore suitable for electrical services in larger buildings and for the interconnection of various buildings. Ahigher DC voltage also offers the following advantages: lower current for same power; high output power lighting; more efficient EV charging points, direct supply of data centers, correct voltage level for inverter driven equipment. The future DC microgrids shall probably consist of a combination of these two voltage levels.The 24V DC bus can be obtained from the 380V DC bus through a high-efficiency
Figure 4:
Interconnected buildings forming a Microgrid cluster.
DC/DC step down converter. The high power and inverter driven equipment would use directly the 380V supply. This concept is shown in Figure 3 where a mix of the two voltage levels could bepresent for different types of equipment. There are already successful implementations of prototype DC microgrids in buildings such as the Next Energy Center, Detriot and the Green. ch-ABB Zurich West Data Centre. In both cases, 380V and 24VDC microgrids are used to provide power for the local equipment. 5 MICROGRID CLUSTERS Microgrids also allow for the formation of clusters (communities) by extending the concept from one building to a group of buildings as shown in Figure 4. This can be carried out both in the case of AC and DC topologies. When considering an interconnected group of buildings, energy management algorithms can be applied to the cluster to enable optimal use of the energy generated from the RES, and increase the efficiency of energy distribution and energy usage.In addition, the microgrid can be designed to beself-sufficient to a certain extentand thus it becomes as independent as possible from the electrical grid. In the case of DC microgrid clusters, such a power system consisting of sources, loads and storage systems can take place without the power quality issues normally present in AC supply systems. DC energy can be shared from one building to another, or stored without too many conversion stages which increases the efficiency of energy distribution. Both DC and AC microgrids haveshorter distribution distances when compared to the conventional power system thus they decrease the distribution losses. Another advantage of microgrid clusters is that multiple buildings can be supported by a common distributed generation source (instead of each building having its own source).
April 2017 ISSUE 56
11
Microgrids – the future of Electrical Energy Grid Continued
6 IEPC RESEARCH ON MICROGRIDS The Department of Industrial Electrical Power Conversion (IEPC) at the Faculty of Engineering of the University of Malta has been actively carrying out research on AC microgrids for a number of years [6], [7], [8], [9], [10], [12] ,[14] and more recently they have commenced work on DC microgrids [11], [13]. The dept. of IEPC has constructedan AC microgrid system which consists of three DG sources connected to a common AC grid via 2kVA inverters. The setup schematic is shown in Figure 5. This setup can work in stand alone or even in grid-connected operation. Both linear and non-linear loads were connected to the AC microgrid to carry out tests to analyse its operational performance.
The knowledge gained by the IEPC in the field of microgrids shall not be limited to power distribution in buildings but shall also be used in another two exciting emerging fields of engineering, namely More Electric Aircraft and Electric Ships. The IEPC also carries out post graduate research degrees (MSc or PhD) in Microgrid engineering and welcomes interested prospective undergraduate/postgraduate students to get in touch with the department.
Experimental tests were carried out to study the laboratory’s microgrid performance during active and/or reactive power demands and verify the load sharing capabilities of the individually controlled inverters [6], [8], [9]. Figure 6 shows the experimental real & reactive power outputs for load demands of 143.5VAr, 71.5VAr and 107.5VAr and 449.3W, 219.3W and 329W per inverter. (The difference in inverter power sharing was purposely introduced for testing purposes.) This setup was also used to carry out tests to verify the seamless transition of the microgrid from standalone to gridconnected and vice versa[10]. Currently the IEPC is also constructing a DC microgrid to continue its research on microgrids. The DC microgrid shall be constructed from a number of DC/DC converters and shall be used to implement advanced energy control algorithms to allow for parallel power sharing and integration of energy storage systems. Figure 7 shows a photo one of the 5kW prototype DC/DC converters constructed at the laboratories for this DC microgrid setup.
Figure 6:
Real and Reactive Power Load Demands inside droopcontrolled AC Microgrid.
Figure 5:
AC Microgrid setup inside the IEPC Energy Laboratory.
12
Figure 7:
DC/DC Converter Unit for future DC Microgrid for the IEPC Energy Laboratory.
7 References
[1]. Bruce Nordman and Ken Christensen, “DC Local Power Distribution,” IEEE Electrification Magazine, June 2016. [2]. Luis Eduardo Zubieta, “Are Microgrids the Future of Energy? ,” IEEE Electrification Magazine, June 2016. [3]. Julio Romero Agüero, Amin Khodaei,and Ralph Masiello, “The Utility and Grid of the Future,” IEEE Power & Energy magazine September/October 2016. [4]. Estefanía Planas, Jon Andreu, José Ignacio Gárate, Iñigo Martínez de Alegría, Edorta Ibarra, “AC and DC technology in microgrids: A review,” Renewable and Sustainable Energy Reviews (2015), pp. 726–749. [5]. Weixing Li, Xiaoming Mou, Yuebin Zhou and Chris Marnay,“On Voltage Standards for DC Home Microgrids Energized by Distributed Sources, 2012 IEEE 7thInternational Power Electronics and Motion Control Conference ECCE Asia June 2-5,2012, Harbin, China [6]. A. Micallef, M. Apap, C. Spiteri Staines and J. M. Guerrero,“ Secondary Control for Reactive Power Sharing in Droop Controlled Islanded Microgrids”, ISIE 2012, 21st IEEE International Symposium on Industtrial Electronics , Hangzhou, China, May 28 -31, 2012. [7]. A. Micallef, M. Apap, C. Spiteri-Staines and J. M. Guerrero,” Cooperative Control with Virtual Selec-tive Harmonic Capacitance for Harmonic Voltage Compensation in Islanded Microgrids”, 38th Annual Conference of the IEEE Industrial Electronics Society IE-CON2012, 25-28 October 2012, Montréal, Canada. [8]. A. Micallef, M. Apap, C. Spiteri Staines and J. M. Guerrero,“ Secondary Control for Reactive Power Sharing and Voltage Amplitude Restoration in Droop Controlled Islanded Microgrids”, 3rd International Symposium on Power Electronics for Distributed Gen-eration Systems(PEDG’12), Aalborg University, Denmark 2012. [9]. A. Micallef, M. Apap, C. Spiteri Staines and J. M. Guerrero,J.C.Vasquez, "Reactive Power Sharing and Voltage Harmonic Distortion Compensation of Droop Controlled Single Phase Islanded Microgrids," Smart Grid, IEEE Transactions on,vol.5, no.3, pp.1149-1158, 2014. [10]. A. Micallef, M. Apap, C. Spiteri Staines and J. M. Guerrero,“Single-Phase Microgrid With Seamless Transition Capabilities Between Modes of Operation,” IEEE Transactions on Smart Grid Vol:6, pp: 2736 –2745, Nov. 2015 [11]. D.Zammit, C. Spiteri Staines, M.Apap, A.Micallef, “Paralleling of Buck Converters for DC Mi-crogrid Operation,” IEEE Conference: International Conference on Control, Decision and Information Tech-nologies (CoDIT 2016), Malta 2016. [12]. A. Micallef, M. Apap, C. Spiteri Staines and J. M. Guerrero,“ Performance comparison for virtual impedance techniques used in droop controlled islanded Microgrids,” International Symposium on Power Electronics, Electrical Drives, Automation and Motion, Speedam, June 22-24, 2016. [13]. D.Zammit, C. Spiteri Staines, M.Apap, A.Micallef, ''Overview of Buck and Boost Converters for DC Microgrid Operation'', Offshore Energy and Storage Symposium (OSES2016), 13th-15th July 2016, Valletta, Malta. [14]. A. Micallef, M. Apap, C. Spiteri-Staines and J. M. Guerrero, "Mitigation of Harmonics in Grid-Connected and Islanded Microgrids Via Virtual Admit tances and Impedances," in IEEE Transactions on Smart Grid, vol. 8, no. 2, pp. 651-661, March 2017.
Prof. Dr Ing. Cyril
Cyril Spiteri Staines graduated with a First Class B.Eng. degree from the University of Malta in 1995 and with a Ph.D. degree from the University of Nottingham in 1998. During 2003-2004 he was a visiting scholar at the School of Electrical and Electronic Engineering at Nottingham. He is a Full Professor with the Department of Ind. Elec. Power Conversion of the Faculty of Engineering of the University of Malta. His current research interests are related to design and controlof Electrical Machines, Power Electronic Converters, Microgrids, Energy Efficiency and Renewable Energy Sources (RES).
Mr Daniel
Zammit
Daniel Zammit received the B.Eng and M.Sc. degrees in Electrical Engineering from the University ofMalta in 2007 and 2012, respec-tively. He is employed as a Systems Engineer with the Department of Ind. Elec. Power Conversion at the Faculty of Engineering, University of Malta. He iscurrently reading for a Ph.D. degree on DC Microgrids. His research interests include Grid Connected Inverters, Power Converters and Microgrids.
Dr Ing. Maurice Dr Alexander
Micallef
Alexander Micallef received the B.Eng.(Hons.),MSc and the PhD degrees in Electrical Engineering from the University of Malta in 2006, 2009 and 2015, respectively. In 2008, he joined the Department of Ind. Elec. Power Conversion at the University of Malta as an Assis-tant Lecturer and in 2015 he became a Lecturer with the same department. In 2012, he was a visiting PhD student at the Department of Energy Technology, Aalborg University, Denmark. His research interests includepower electronic converters, renewable energy systems, control and management of distributed generation and energy storage systems, smart grids, and microgrids.
Spiteri Staines
Apap
Maurice Apap received the B.Eng (Hons.) and MSc degrees in Electri-cal Engineering from the University of Malta in 1996 and 2001 respectively and the PhD degree from the University of Nottingham in 2006. He worked at ST Micro electronics from 1996 to 1998 before resuming his studies and also worked as a researcher on matrix converter drives for aerospace projects with the Univer-sity of Nottingham during and following his PhD studies. He is currently a senior lecturer at the Department of Ind. Elec. Power Conversion at the University of Malta. His research interests include power electronic converters, microgrids and the control of electrical drives.
April 2017 ISSUE 56
13
at BOV your wealth is in safe hands Whether you are a conservative, balanced or growth orientated investor, our qualified financial advisors located at the Bank’s Investment Centres can provide you with tailor-made financial planning solutions. Visit any BOV Branch or Investment Centre in Malta or Gozo to discuss your financial planning requirements and to set an appointment with one of our professional Financial Advisors.
your success is our goal
BOV INVESTMENT CENTRES
2131 2020 I bov.com
Bank of Valletta p.l.c. is licensed to conduct investment services business by the Malta Financial Services Authority.
Issued by Bank of Valletta p.l.c., 58, Triq San Ĺťakkarija, il-Belt Valletta VLT 1130
BOV Investment Centres – The Balancing Act
Inauguragted in 2012, the BOV Investment Centres set out to be ‘Centres of Expertise’ bridging the gap between the Bank’s Wealth Management arm and the investment services offered from its branch network. Five years after the six centres were inaugurated, we take a closer look at what these Centres are and how they work. Bank of Valletta understands that all customers expect their advisor to manage their wealth professionally as they would manage their own business, irrespective of how affluent they are. Therefore they expect efficiency and competence, at a competitive cost, along with peace of mind. In order to offer this type of service to non-Wealth Management customers, at affordable rates, the Bank came up with the concept of Investment Centres. However, prior to taking that step, it underwent an in-depth exercise, during which it recategorised its investment products according to their risk levels and volatility. Only then, did it transfer its investment advisory services from the retail network to its Investment Centres. The typical BOV Investment Centre is composed of six to eight Financial Advisors and one Portfolio Administrator led by a manager who is responsible for the general overseeing of the Centre, ensuring smooth operation at all times. The Financial Advisors are fully qualified and authorised by the MFSA to provide investment advice. They are well versed and keep themselves constantly updated, not only on the extensive range of investments they offer, but also on the macroeconomic environment, which could have a major impact on the performance of the investments and on the level of financial advice that they provide their clients. The Centres do not work in isolation. They are supported by a specialised unit that feeds them regular market updates and training, as well as delivering independent, accurate and unbiased research and analysis in a timely manner. Every Financial Advisor is responsible for a number of relationships with whom s/he makes it a point to meet on a regular basis, in line with the clients’ exigencies. Regular discussions with clients enable Financial Advisors to better understand the requirements of each and every individual client, in order to be in an optimal position to identify the right products that match the client’s needs in line with the established risk profile.
Left to right:
Josianne Theuma Dimech, Sandra Magro, Rose La Rosa,
Aldo Scardino, Roberta Bellizzi, Claire Falzon and Ethelbert Perini
Ever since these Centres were inaugurated, the services offered and performance was monitored through regular customers surveys conducted amongst other channels. Feedback shows that clients have embraced the Investment Centre concept, and are satisfied with the holistic experience they recieve when visiting the Centres. In this regard, they refer specifically to the high level of privacy and confidentiality which they are provided within these Centres. Reactions are also positive with respect to the timeliness of appointments and the execution timeframes within which, instructions received from clients are executed. These regional centres have succeeded in becoming an important link in the Bank’s investment offering. They live up to the Bank’s trademark, with the Financial Advisors making it a point to build a strong working relationship with each and every one of their clients. This enables them to customise their offering to the client’s particular circumstances, every time. By retaining an active and open working relationship, the service remains personalised and alive, changing continuously to reflect the changing circumstances and risk profile of the client. If you are interested in learning more about how we can assist you with your investments, whether you already have investments, or are seeking advice for the first time we invite you to contact us on 2131 2020, so we can discuss your investment requirements at your convenience.
The opinions expressed herein should not be interpreted as investment advice. Past performance is not a guarantee of future performance. Investments should be based on the full details of the Prospectus, Offering Supplement and the KIID which may be obtained from BOV Asset Management Limited, Bank of Valletta p.l.c. Branches, Investment Centres and other Licensed Financial Intermediaries. BOV Asset Management is licensed to provide Investment Services in Malta by the MFSA. Issued by BOV Asset Management Limited, Registered address: 58, Triq San Żakkarija, Il-Belt Valletta VLT 1130
EMG Signal Analysis and Finger Movement Classification Adrian von Brockdorff adrian.brockdorff@gmail.com
ABSTRACT This study covers the analysis of electromyographic (EMG) signals, which are the electrical signals created during muscle contraction. It also covers the effect of various noise and artifacts that may interfere with the raw signals and the implementation of a cost-effective and simple to use EMG device that can classify different finger movements. Two methods were tested to evaluate different types of EMG signal acquisition techniques. The first method consisted of placing five electrodes on the muscles which are most active when the respective finger is contracted, while the second method consisted of placing six surface electrodes at the lower part of the forearm. Experimental tests showed that the second method provided the most repeatable and accurate results. This study also covers how a cost effective EMG signal acquisition module can be used to control any external devices, such as prosthetic hands or an electric wheelchair. This is achieved by minimizing the number of components and using efficient algorithms to reduce the power consumption and the cost of the microcontroller. Keywords: EMG, biomedical engineering, pros-thetic hands, finger classification.
16
1 INTRODUCTION EMG stands for electromyography, which is the study of muscle function through analysis of the electrical signals emanated during skeletal muscular contraction. EMG signals are based upon action potentials at the muscle fibre membrane resulting from depolarization and repolarization processes. Muscle fibre contracts when an action potential is produced along an alpha motor neuron. These action potentials are measured using EMG equipment. The typical amplitude of the raw signal can vary between 0mV to 10mV peak-to-peak at a frequency of 5Hz to 450Hz. For this study, the raw EMG signals were detected using surface electrodes on the forearm. Following the acquisition, the signal was filtered to leave the dominant energy and attenuate any noise and artifacts that may deter the quality of the signal processing. Different EMG signal characteristics were studied and analysed to determine the best algorithm to be used for obtaining different signal classifications. These include the effects of the signal amplitude and frequency with finger force, angle and fatigue. The repeatability of the signal characteristics were also studied to determine on which parameter the algorithm was to be based. Following the study of the raw signals and their characteristics, two different surface electrodes placements were tested to determine which method provides the most reliable solution. The first experiment consisted of placing five electrodes on the muscles which are most active when the respective finger is contracted. The second experiment made use of six electrodes which were placed on the lower part of the forearm.
ring finger is contracted. Figure 1 shows the placement of the electrodes. All signals were examined in both time and frequency to determine the ideal domain to be used from the processing. 2.1 Analysis of baseline noise and movement artifacts The baseline noise and movement artifacts were the first characteristics that were studied on the raw EMG. With the arm at rest, the baseline noise contribution was of around 10mVp-p (Figure 2). Although the arm was at rest and the electrodes were static, a slight baseline drift was still observed in the signal. When the signal was analysed in the frequency domain, a high noise floor was noticed at all the frequency ranges that were of interest to the study, with the lower frequencies (0Hz to 5Hz) having a higher magnitude due to baseline wander. To further analyse the baseline wander effect on the raw signal, the electrodes were gently moved in different directions to simulate a movement of the arm. Figure 3 shows the results obtained from this test. The results show that the noise contribution remained the same, but the baseline drift increased significantly due to the electrodes being moved. This movement in the signal baseline would make it difficult to analyse and process the signal in both the time and frequency domain. This would also contribute to complications in classifying individual finger movements. Through frequency domain analysis it was noted that this drift was at the low frequency spectrum ranging from 0Hz up to 5Hz.
The method which gave the best results was selected and their algorithm and hardware parts were implemented. A pneumatic hand was also constructed to mimic the hand movements. 2 ANALYSIS OF THE RAW EMG SIGNAL For this part of the study, a Data acquisition (DAQ) instrument was used to log the EMG signals acquired from the surface of the forearm. One pair of electrodes was placed over the Palmaris Longus muscle, which is most active when the
Figure 1:
Electrodes placement for analysing noise and movement artifacts during the ring finger contraction.
Figure 2:
EMG noise (top) and drift (bottom).
April 2017 ISSUE 56
17
EMG Signal Analysis and Finger Movement Classification Continued
2.2 Analysis of EMG burst repeatedly This test consisted of repeatedly contracting the ring finger to check whether the EMG signal amplitudes were constant with every finger contraction. This was an important test to determine whether the EMG signal in the time domain would be reliable enough for classifying finger movements. The ring finger was closed repetitively, with the results obtained illustrated in Figure 3. This test indicated that the EMG bursts were quite similar in amplitude. To further analyse the amplitudes, the raw EMG bursts were enveloped using the moving root-mean-square (RMS) filter. This process consisted of three stages. In the first stage, the EMG signal data points were squared. This was required to rectify the raw signal. Following this, the average of the squared data points was taken to envelope the signal. The last stage consisted of taking the square root of the enveloped signal to obtain the original amplitude level. Two different averaging window widths were tested for the RMS algorithm, with the first window width being 1000 samples and the second being 3000 samples (see Figure 4). It was observed that with an RMS of a window width of 1000 samples, the output was not smooth enough to deduce a clear interpretation for the amplitude of the EMG burst. When the width was increased to 3000 samples, noticeable smoothing of the EMG enveloping was noticed. The difference
Figure 3:
between the lowest and the highest RMS amplitude was very slight, with the lowest being around 0.115V and the highest being around 0.125V. This is just a 10mV difference, which when compared to the 3.3V ADC voltage reference is insignificant (0.3%). This difference can be further reduced if the averaging window width is increased. This is due to the fact that more samples will be averaged and thus a smoother enveloping of the signal will be formed. Another advantage of having a wide averaging window in the RMS algorithm is that any unwanted spikes in the raw EMG will not affect the quality of the overall signal. The only drawback to having a wide averaging window is that the output will be slightly delayed. Therefore, one must compromise between the two to get a smooth enveloping with the least possible output delay. Other experiments on the EMG signals were conducted such as, the analysis on the change in amplitude and frequency with muscle fatigue and the change to the raw EMG signal with different forces. From these experiments it was concluded that the best method would be to process the signals in the time domain rather than the frequency domain since the readings show more similar receptiveness which is more significant for the purpose of this project. 3 FINGER CLASIFICATION EXPERIMENTS 3.1 First experiment with electrodes placed on the respective finger muscles The first experiment for obtaining readings from individual finger movements consisted of placing the electrodes on the muscles which are most active when a particular finger is contracted. Figure 5 illustrates the placement of the five electrodes, where each electrode was used to detect the contraction of each finger. Every electrode in Figure 5 represents a finger as follows: • Electrode 1 for the index finger • Electrode 2 for the ring finger • Electrode 3 for the thumb • Electrode 4 for the little finger • Electrode 5 for the middle finger
Raw EMG burst repeatedly.
Figure 4:
RMS of EMG bursts with a window width on 3000.
Figure 5:
Electrodes Placements.
April 2017 ISSUE 56
19
EMG Signal Analysis and Finger Movement Classification Continued
With the system power-on, a calibration session was performed so that the system could set the thresholds for each electrode. The user was instructed to close one finger at a time, during which the system monitored the changes in the amplitude values of the respective electrode. The system then stored the highest recorded amplitude in a variable. This process was repeated for a pre-defined amount of times, which could be varied via the program. After the test was repeated for a number of pre-defined repetitions, the program took the average of all recorded amplitudes and set upper and lower boundary limits (Figure 6). During normal operation, the system continuously monitored the amplitudes of the respective electrode and compared each sample with the thresholds set during the calibration process. If the input sample lays between the upper and lower thresholds, then the system will assume that the respective finger is contracted. After testing this method on different participants, it was concluded that not all the outputs obtained were detecting the correct finger movements. In fact, two out of four participants had at least one different finger classification from what was expected. These discrepancies in the results may have been caused by muscle anatomy which varies from one person to another. A drawback which was also noticed when using this EMG acquisition method is that some of the electrodes were placed very close to muscles which are used to flex and extend the wrist and the elbow. This resulted in cross-talk which caused the EMG amplitude to exceed the threshold and trigger the outputs unintentionally. The fact that two of the electrodes were placed at the upper part of the forearm were also causing inaccurate readings for overweight participants because fat increases the impedance between the muscles and the electrodes. The electrodes placed on the lower part of the forearm were not affected by this problem.
Figure 7:
Electrodes placement for the second experiment.
Figure 8:
A and B - all muscles in the forearm, C - flexor digitorum superficialis muscle, D - flexor digitorum profundus muscle, E - flexor pollicis longus muscle.
3.2 Second experiment with elec-trodes placed on the lower part of the forearm Since the first experiment did not yield satisfactory results, a second experiment was conducted. In this experiment, six electrodes were placed on the lower part of the forearm as shown in Figure 7. However, there was no need for the
Figure 6:
Graphical representation of the calibration process used for the first experiment.
20
Figure 9:
Setup used for the second experiment.
electrodes to be placed precisely on the areas indicated in Figure 7, since the algorithm used for this experiment does not rely on the EMG signals emitted from the belly of the muscles. Instead, this algorithm relies on all the EMG signals acquired from all six electrodes during a particular finger movement. This was an advantage over the first experiment where all electrodes had to be placed at the precise area with no tolerance for misalignment.
Figure 10:
Plotting of all the amplitudes and the thresholds set when the little finger was contracted.
Figure 11:
Plotting of all the amplitudes and the thresholds set when the little finger was contracted.
The forearm has nineteen major muscles which are responsible for flexion, extension and other movements of the fingers, wrist and elbow. After studying the anatomy of the muscles, it was concluded that the muscles used for the contraction of the fingers are mostly exposed at the lower part of the forearm. Hence, placing the electrodes in this area drastically reduces cross-talk from other muscles which were not intended to be monitored for this project. Figure 8 shows the all muscles found in the forearm and the main three muscles that are responsible for finger movement. These are the flexor digitorum superficialis (responsible for flexing all fingers - primarily at proximal interphalangeal joints), the flexor digitorum profundus (responsible for flexing the distal and proximal interphalangeal joints) and the flexor pollicis longus (responsible for flexing the thumb). The way the electrodes were placed for this experiment is not common practice. In fact, most researchers place the electrodes around the upper part of the forearm. This method of placing the electrodes on the upper part of the forearm has both positive and negative aspects. The advantage is that electrodes in this area will pick EMG signals directly from the belly of the muscles, thus the signal will be less sensitive to noise distortion and so less amplification will be required. The drawback of using this method is that algorithms which only detect the required muscle movements are very complex since the algorithm has to identify which EMG signals are emitted from the muscles related to finger movements and attenuate any signals coming from other muscles in the forearm. Using complex algorithms requires specific processors that can process intricate digital signal processing. Another drawback is that if the learning neural network is not well designed, unwanted muscle activities can be represented as wanted signals. One would also have to use at least ten electrodes to get the required data for the neural network to work correctly. The decision to place the electrodes on the lower part of the forearm for this experiment was taken to test the performance and accuracy of a new technique and to eliminate the use of complex algorithms which require intricate programming structures. After placing the electrodes, a calibration process had to be followed. This process consisted of contracting each finger multiple times. The user was instructed to close the fingers one at a time, and with each contraction the amplitudes acquired from all electrodes were monitored and stored. This process was repeated for a pre-defined amount of times so that the system could determine the required thresholds. The monitored EMG bursts were then used to form the upper and lower thresholds for each finger contraction, with the highest monitored amplitude being set as the upper threshold and the lowest amplitude being set as the lower threshold. Figure 9 illustrates the setup used for this experiment, while Figure 10 shows the plots obtained for the little finger only.
Figure 12:
Complete setup.
From the data obtained, it was noted that the thresholds for each finger has a unique pattern. These patterns were utilized
April 2017 ISSUE 56
21
Developing innovative elevators that’s what inspires us.
www.thyssenkrupp-elevator.com Increased transport capacity Reduced space requirements Enhanced safety Improved energy efficiency Greater durability and recyclability Residential and Office Buildings | Shopping Centres and Hotels | Hospitals and Airports
ELEVATORS FOR BUILDINGS OF ALL KINDS
AR Tech, Fleur De Lys Road, Birkirkara, Malta Phone: +356 2144 1300 | Fax: +356 2149 8474 | sales@artechmalta.com
EMG Signal Analysis and Finger Movement Classification Continued
for identifying different finger contractions. This is done by constantly monitoring the amplitudes from all electrodes and whenever they lay between these unique thresholds, the algorithm would detect the similarity and activate the respective output. This experiment yielded a repeatability of 92% and an accuracy of 100%, with results taken from 40 contractions from four different participants. 4 HARDWARE IMPLEMENTATION Following the satisfactory outcome from the second experiment, the hardware was converted from the prototype boards to the actual PCB implementation. The hardware consisted of six active electrodes, a signal conditioning module and the MCU main module. A prostatic hand was also constructed to replicate the finger movements. Figures 11 and 12 shows the final hardware implementation.
Mr Adrian
von Brockdorff
Adrian von Brockdorff graduated from MCAST with a degree in Electronics Engineering. He is currently working at STMicroelectronics as a Maintenance Engineer. His main areas of interest are electronics and robotics.
April 2017 ISSUE 56
23
Proportional and Simultaneous Myoelectric Control of a Robotic Arm Christian Grech, Tracey Camilleri, Marvin Bugeja Department of Systems and Control Engineering, University of Malta christian.grech.12@um.edu.mt
ABSTRACT In the past years, machine and robot technologies have progressed rapidly, improving human life and making tasks much easier to execute. Human-machine interfaces (HMI) are a vital tool for people with mobility impairments, giving them the possibility of controlling a machine remotely. This project's aim was to develop a HMI for the control of a robotic arm manipulator, through the use of surface electromyography (EMG) signals. A model which converts EMG signals from multiple muscles to elbow and shoulder angles for simultaneous and proportional control was developed and tested in real-time. This model was applied to seven different single joint and simultaneous movements in different 2D planes and a real-time application was developed. Keywords: Electromyography (EMG), linear regression, multi-layer perceptron, simultaneous proportional control, state space model.
24
1 INTRODUCTION Biosignals are becoming increasingly important in many practical uses including biomedical and clinical applications, prosthetic devices, human machine interfaces (HMIs) and more. HMIs are mostly used to help people with reduced mobility to control a machine remotely. One possible control interface is through biosignals which are continuous signals recorded from the human being. This project focuses on the use of muscle activity, known as electromyography (EMG), for the control of a HMI system, by making use of a model which converts EMG signals to joint angles. Furthermore, the ability of the model to control more than one degree of freedom in a natural manner will be evaluated. This type of control is known as proportional and simultaneous myoelectric control. In this project, the aim was to compare different EMG-tojoint angle models and identify the one which can replicate movements in different planes with minimal error. Specifically, state space models, linear regression models and a multilayer perceptron neural network were compared on 2D elbow movements and the latter, which was found to give the lowest root mean square error, was tested on other single and simultaneous shoulder and elbow movements. Finally, a real time implementation of the system was developed, in which a robotic arm was made to mimic the subject’s upper arm movement. 2 EMG-TO-JOINT ANGLE MODELS This project focuses on the conversion of EMG data directly to output joint angles. The three types of models used are detailed below. 2.1 State space representation The state space representation is a method used in modern control theory to avoid representing complex systems using transfer functions. Various researchers have made use of this model for EMG-to-joint angle translation. Artemiadis and Kyriakopoulos make use of this model in [1] where EMG signals from nine electrodes on the upper limb muscles are used to control a robot arm with four degrees of freedom. Artemiadis and Kyriakopolous argue that a model which describes the function of the muscles in simulating the human joints would be generally complex, making it difficult to implement real-time decoding. For this reason, the state space model is used as it is a flexible decoding model in which
hidden variables could model the unobserved system states and translate arm motions from EMG signals. 2.2 Linear Regression Linear regression (LR) model structures provide a useful method in describing basic linear systems. This method attempts to model the relationship between two variables by fitting a linear equation to the data. One method which is commonly used to fit a regression line is the least meansquares method. The best fitting line is found by minimizing the sum of squares of the vertical deviations from each point to the regres-sion line. The benefit of this technique is that provides a continuous output value. In fact, Hahne et al. in [2] implement a linear regression model where the instantaneous feedback of the closed-loop system is exploited such that the device can improve the joint position accuracy on the fly. 2.3 Multilayer Perceptron Another form by which the EMG-to-joint angle model can be implemented is by making use of artificial neural networks (ANNs). An ANN is made up of a considerable amount of simple processing units, coupled by weighted connections, where each unit receives inputs from several other units and produces a single output. This then acts as an input to other units. A multi-layer perceptron (MLP) is one such type of neural network made up of simple neurons called perceptrons. A MLP network has an input layer (on the left), hidden layers (in the middle) and an output layer (on the right) as can be seen in Figure 2. Once the architecture to be used is chosen, the network's weights and thresholds are set so as to minimise the prediction error. A back-propagation neural network (BPNN) algorithm can be used to train MLPs. The nature of this algorithm is to minimise the error of the network using the derivatives of the error function. The advantage of using a neural network is its ability to represent both linear and non-linear relationships, and learn these relationships directly from the data being modelled. In [3], Muceli and Farina use eight MLPs to estimate several hand movements. Different MLPs were used for different wrist and hand angles with the aim of obtaining an accurate estimation for each degree of freedom. Similarly, in [4], Aung and AlJumaily used a neural network to estimate the shoulder and elbow joint angles from the recorded EMG signals.
Figure 1:
Results obtained in this project.
April 2017 ISSUE 56
25
Natura Collection HANDCRAFTED
INSPIRATIONS
Valletta Glass is an innovative design based studio with an unrivalled reputation in standard and bespoke art glass products. It’s Creations are the result of Mediterranean design coupled with traditional European glassblowing and fusion techniques.
Trophies ∙ Memorabilia ∙ Office Decor Home Decor · Collections · Tableware Jewellery · Fusion info@vallettaglass.com
Tel: 2143 0966
Ta’ Qali Crafts Village, Long Hut One
TA’ QALI MDINA VALLETTA BUGIBBA
ABERTAX® – Leaders in Battery Ancillaries
Abertax innovative products will ensure a quality battery service www.abertax.com ABERTAX TECHNOLOGIES Ltd., KW17A Corradino Industrial Estate, Paola PLA 3000, Malta, Tel +356 23 678100, +356 21 660606, Fax +356 21 808625
Patent protection on all products and concepts
Abertax – 25 years experience with High-Tech Battery Ancillaries. World leader with the development and production of Battery Management Units (BMU), Battery Monitoring Systems (BMS), Capacitive electrolyte Level Sensors (CLS), Gas Release Systems for VRLA Batteries (GRS) , Intelligent Battery Concepts (IB) and Washer with Leading Wire (WLW).
Proportional and Simultaneous Myoelectric Control of a Robotic Arm Continued
3 SYSTEM IDENTIFICATION 3.1 Procedure The three EMG-joint angle modelling techniques identified in the previous section were analysed by estimating the elbow joint an-gle in the x-z plane as can be seen in Figure 1a. Using the Vicon motion capture system [5], ten trials, each 40 seconds long, were recorded with constant-speed arm movement. The EMG signals from the muscles were recorded using the ZeroWire acquisition box [6]. Three markers were used to determine the arm's position: at the shoulder, elbow and wrist as can be seen in Figure 4. Inverse kinematic equations were then applied to these three co-ordinates so as to find the angle of the elbow. Analysis of each model was carried out by testing on cross-validated data, allowing the possibility to analyse the model’s ability to generalize to new data. The performance measures used were the root mean square error (RMSE), the correlation coefficient (CC) and the coefficient of determination (R2 ). The numerical subspace state space system iden-tification algorithm (N4SID) was used to estimate the state space model’s parameters. For the re-maining two models, the input EMG signals were preprocessed by applying the root mean square (RMS) feature on the rectified EMG signal, with a window size of 200 samples. Furthermore, the data was filtered using a 1Hz lowpass filter and normalized. In the case of the LR model, MATLAB's Statistics and Machine Learning Toolbox was used, specifically the film, to construct a least-squares fit of a model to the input data. On the other hand, MATLAB’s Neural Networks Toolbox was used to represent and train the MLP network. The backpropagation al-gorithm used to train the network was the Le-venberg-Marquardt algorithm as suggested by Aung and Al-Jumaily [4]. As can be seen in Figure 2, the MLP architecture consisted of an input layer with two input nodes, one hidden layer with three sigmoidal nodes, and an output layer with one sigmoidal node. The number of nodes in the hidden layer was optimized with cross-validation.
3.2 Results and Discussion The results obtained for elbow movements in the x-z plane with the three different models being compared are shown in the first three rows of Table 1 while Figure 3 compares their performance according to the RMSE and CC values. Figure 3a shows that in terms of the RMSE, the MLP performed superiorly to the other methods, and also more consistent,
Figure 2:
MLP structure for elbow joint angle estimation.
Figure 3:
Performance across different models.
Joint
Plane
Method
RMSE(°)
CC
Elbow
x-z
State space
26.73 ± 5.80
0.690 ± 0.148
Elbow
x-z
LR
31.65 ± 19.22
0.869 ± 0.063
Elbow
x-z
MLP
15.16 ± 4.22
0.914 ± 0.020
0.736 ± 0.143
Shoulder
x-z
MLP
12.34 ± 1.43
0.936 ± 0.018
0.868 ± 0.035
Shoulder
x-z
MLP
8.81 ± 1.61
0.910 ± 0.018
0.736 ± 0.086
Shoulder
y-z
MLP
5.43 ± 0.33 Elbow
Shoulder and Elbow
x-z
MLP
Shoulder and Elbow
x-y
MLP
Shoulder and Elbow
y-z
MLP
R2
0.970 ± 0.002
Shoulder
Elbow
Shoulder
N/A
0.932 ± 0.008 Elbow
Shoulder
9.03
7.30
0.977
0.943
0.945
0.823
± 0.75
± 0.85
± 0.003
± 0.009
± 0.018
± 0.074
11.92
11.94
0.956
0.894
0.837
0.316
± 2.84
± 2.54
± 0.003
± 0.020
± 0.010
± 0.291
25.46
9.53
0.862
0.938
0.455
0.588
± 6.12
± 1.16
± 0.039
± 0.011
±0.131
± 0.176
Table 1:
Results obtained in this project.
April 2017 ISSUE 56
27
Proportional and Simultaneous Myoelectric Control of a Robotic Arm Continued
with the lowest standard deviation. The second graph, Figure 3b shows the CC performance for the three models, where the MLP model performs better on average than the linear regression model. This outcome shows that using a non-linear method such as the MLP is a better option for estimating the joint angle using EMG signals. 4 MODELLING DIFFERENT MOVEMENTS This comparative analysis has shown that the multi-layer perceptron (MLP) approach gives the best results in terms of estimated angular error. Hence, this approach was used to model other single and simultaneous shoulder and elbow joint angle movements as detailed below. 4.1 Single joint movements The first three modelled movements were single joint movements in three different planes. These included shoulder movement in the x-z plane, x-y plane and the y-z plane. Ten trials, each 40s long, were recorded for each movement. The Vicon motion capture system was used to record the position of three markers so that the joint angle can be calculated. For the x-z plane shoulder movement seen in Figure 1b, the same MLP structure as that in Figure 2 was used, however instead of using the biceps brachii and triceps brachii as the pair of input EMG channels, the anterior and posterior deltoid channels were used. The output in this case was the shoulder angle. The three position markers were placed on the shoulder, elbow and on the hip as can be seen in Figure 4.
EMG signals. For every single joint movement, the root mean square (RMS) of the rectified EMG signals was used to train the network. The average cross-validated results obtained in the three tests can be seen in Table 1. Comparing all the single joint movements covered in this project, as shown in Figure 5, one can note that there was a slight difference in CC values but quite a notable difference in the RMSE values. With an average RMSE of 5.43° with cross-validation, the shoulder movement in the y-z plane was the best performing single joint model in this project. 4.2 Simultaneous Joint movements Following the modelling of four single joint movements, three simultaneous shoulder and elbow movements were modelled in three different planes using the MLP approach. The first movement, simultaneous elbow and shoulder movement in the x-z plane combines the elbow flexion/extension and the shoulder flexion/extension movements, as can be seen in Figure 1e. To capture this movement, both the elbow and
The next modelled movement was in the x-y plane as can be seen in Figure 1c. In the training session, the three position markers were placed on the shoulder, elbow and on the chest. Two EMG probes were used to record signals at the posterior deltoid and pectoralis major, and used as inputs to the twoinput MLP structure. The final modelled movement was in the plane as shown in Figure 1d. For the training session, the three position markers in this case were placed on the shoulder, elbow and on the hip. A two-input MLP structure was used, the inputs being the anterior and posterior deltoid
Figure 4:
Marker and EMG electrode placement.
28
Figure 5:
Performance of sequential joint movements in terms of CC, R2 and RMSE.
shoulder angle were recorded simultaneously, together with four EMG channels: biceps brachii, triceps brachii, anterior deltoid and posterior deltoid. In this case, the MLP structure consisted of four input nodes, three neurons in the hidden layer and two output nodes. The second modelled movement was simultaneous elbow and shoulder movement in the x-y plane, as can be seen in Figure 1f. To capture this movement, both the elbow and shoulder angle were recorded simultaneously, together with three EMG channels: biceps brachii, triceps brachii and posterior deltoid. For this model, the same MLP structure as in the previous case was used, however with three input nodes, three neurons in the hidden layer and two output nodes. The final modelled movement, simultaneous elbow and shoulder movement in the y-z plane can be seen in Figure 1g. Three EMG channels were used: biceps brachii, triceps brachii and anterior deltoid. As in the last movement, the MLP structure in Figure 2 was used, with three input nodes, three neurons in the hidden layer and two output nodes. For all of the movements, after
pre-processing the EMG data using the RMS feature, the weights were calculated and the model was tested with crossvalidation. The RMSE, CC and R2 parameters were used once again to evaluate the performance of the models. The results obtained can be seen in Table 1. Comparing the three simultaneous movements, Figure 6 shows the elbow and shoulder angle estimation performance for the three simultaneous movements. In this case the simultaneous movement in the x-z plane was the best performing movement with the lowest errors for both the shoulder and elbow estimation. The results obtained in this project were satisfactory, however direct comparison with literature was not possible as most of the previous works either rely on discrete gesture classification techniques or else they use proportional techniques on estimating the wrist angles [3, 7], rather than shoulder and elbow angles as considered in this project. 5. REAL-TIME OPERATION Following the offline analysis, the next goal of the project was to implement the system for real time operation where the subject’s EMG signals are used to control a robotic arm. Training of the MLP neural network was carried out as detailed in Section 4. In the online case, EMG data was then acquired from the user through the ZeroWire EMG acquisition box and transmitted over TCP (Transport Control Protocol). A C++ program automatically configures and activates the EMG acquisition box and also caters for the transfer of this data to MATLAB. The signal is then processed and undergoes the conversion from EMG amplitude to the joint angle through the previously trained model. The estimated angles are then transferred to a Simulink model which controls the robotic arm's movement. The CRS Catalyst-5T, which is a five degree-of-freedom robot system, was used in this project to replicate the shoulder and elbow user movements. As the robotic arm has an elbowup configuration, while the human arm has an opposite configuration, it was not possible to control the robotic
Figure 6:
Performance of simultaneous joint movements in terms of CC, R2 and RMSE.
Figure 7:
Real time wrist position control.
April 2017 ISSUE 56
29
Proportional and Simultaneous Myoelectric Control of a Robotic Arm Continued
arm's angles and achieve an exact replication of more than one degree of freedom. Hence, for simultaneous elbow and shoulder angle replication, it was decided that the position of the subject's wrist will be mimicked by the robotic arm's end effector instead. In order to evaluate the real-time simultaneous movement replication performance, the simultaneous elbow and shoulder flexion/extension movement in the x-y plane was adapted to a typical movement performed every day. Three objects were placed on a horizontal plane, and the robotic arm followed the position of the subject's wrist when it moved and stopped to point at an object. As can be seen in Figure 7, the robotic arm was placed in front of the subject and it was configured so that it performs the mirrored position of the subject's wrist. Comparing the performance obtained to that achieved by Artemiadis and Kyriakopoulos [1] when using a state space model, results were very similar with an average CC of 0.93 for both the x and y directions. 6 CONCLUSION The objective of this project was to implement a system that allows for the continuous, simultaneous control of various degrees of freedom of a robotic arm in real-time using EMG signals from different muscles. Through a literature review, three EMG-joint angle models were identified as suitable options for this project. Through several tests, the MLP was found to be the best performing model, and hence this model was extended to model different single and simultaneous elbow and shoulder movements in different 2D planes. Finally, a real time application was implemented to show how external devices can be controlled directly using muscle activity. Considering the developments achieved in this project, the system replicates natural simultaneous movements using minimal electrodes and computationally inexpensive software which is robust to EMG transient effects when operating in short periods. Since the submission of this final year project, the work has been extended to compare different types of neural networks and evaluate their performance on 5 different subjects. This work has been submitted as a journal publication and is currently under review. 6.1 References
[1] P. Artemiadis and K. Kyriakopoulos, “EMG-Based Control of a Robot Arm Using Low-Dimensional Embeddings”, Robotics, IEEE Transactions on, vol. 26, pp. 393-398, April 2010. [2] J. Hahne, K. Muller, and D. Farina, “An Embedded System for Simultaneous and Proportional Myoelectric Control of Upper Limb Prostheses”, in IEEE Engineering in Medicine and Biology Society, 37th Annual International conference, August 2015. [3] S. Muceli and D. Farina, “Simultaneous and Proportional Estimation of Hand Kinematics from EMG during Mirrored Movements at Multiple Degrees-ofFreedom”, Neural Systems and Rehabilitation Engineering, IEEE Transactions on, vol. 20, pp. 371-378, May 2012. [4] Y. Aung and A. Al-Jumaily, “Estimation of Upper Limb Joint Angle Using Surface EMG Signal," International Journal of Advanced Robotic Systems, vol. 10, no. 369, pp. 1-8, 2013. [5] Vicon [Online]. Available: http://www.vicon.com/ what-is-motion-capture [Cited: 2016-05-09]. [6] Aurion [Online]. Available:
http://www.aurion.it/ZeroWire Technical Features rev02.pdf [Cited: 2016-05-09]. [7] J. Hahne, F. Biebmann, N. Jiang, H. Rehbaum, D. Farina, F. Meinecke, K. R. Muller, and L. Parra, “Linear and Nonlinear Regression Techniques for Simultaneous and Proportional Myoelectric Control”, Neural Systems and Rehabilitation Engineering, IEEE Transactions on, vol. 22, pp. 269-279, March 2014
Mr Christian
Grech
Christian Grech received the B.Eng.(Hons) degree in electrical and electronics engineering from the University of Malta, in 2016. He is currently working toward a Master degree at the Department of Microelectronics and Nanoelectronics at the University of Malta. His main research interests concern signal processing, system identification and modelling.
Dr Tracey
Camilleri
Tracey Camilleri graduated with a B.Eng.(Hons) degree in electrical engineering from the University of Malta in 2004 and completed her PhD in the field of biomedical signal processing from the same university in 2012. She is currently a lecturer with the Department of Systems and Control Engineering at the University of Malta and her research interests include system identification, biosignal analysis and human machine interfaces.
Dr Ing. Marvin
K. Bugeja
Marvin Bugeja received the B.Eng. (Hons.) degree (summa cum laude) in electrical and electronics engineering from the University of Malta in 2002, and completed his PhD in the fields of Computational Intelligence and Robotics at the same university in 2011. He is currently a lecturer with the Department of Systems and Control Engineering at the University of Malta and his research interests include robotics, computational intelligence, and adaptive and nonlinear control.
April 2017 ISSUE 56
31
Sensorless Position Control of a PMSM for Steer-by-Wire Applications Kris Scicluna(1)(2), Cyril Spiteri Staines(1), Reiko Raute(1) (1) Department of Industrial Electrical Power Conversion, Faculty of Engineering, University of Malta. (2) Institute of Engineering and Transport, Malta College for Arts, Science and Technology. Kris.Scicluna@mcast.edu.mt.
ABSTRACT This article presents the design and implementation of a MATLAB/Simulink model for the sensorless control of a PMSM in a steer-by-wire application. Simulation results for position, speed and current loops in a closed-loop sensorless mode are shown. The sensorless method is based on the tracking of saliencies by high frequency injection. Keywords: steer-by-wire; sensorless; PMSM, high frequency injection.
32
1 INTRODUCTION Research on by-wire technologies for automotive applications has increased during re-cent years due to the number of advantages which such systems have over conventional mechanical systems. Such advantages include better fuel economy and reduced gas emissions in throttle-by-wire, reduced risk of skidding in brake-by-wire and reduced volume required for the steering arrange-ment in steer-by-wire [1]. All by-wire systems can also be adjusted for improved vehicle dynamics and driver comfort. The main aim of these by-wire sys-tems is to have driver inputs which are not mechan-ically coupled to the drivetrain of the vehicle but measured through sensors and driven electrically through the use of appropriate drive mechanisms [2].
suffer during low speeds and low operational frequencies [4], a high frequen-cy based sensorless system was implemented.
The most challenging by-wire system to implement is steerby-wire; this is due to the vari-ous sensors and actuators which need to be inte-grated in the system and due to the critical safety associated with the application. In order to facilitate the widespread use of steer-by-wire applications researchers must ensure that the resultant systems are both fault tolerant and with a response which is comparable to traditional power steering arrangements which are already available on the market [3].
Most of the electrical drives used in steer-by-wire applications presented in literature are implemented using PMSMs as shown in Figure 2 [5, 6], due to a number of advantages associated with this machine such as: higher effi-ciency and higher power density compared to brushed DC drives, better heat dissipation charac-teristics compared to induction machine based drives, no commutator maintenance and higher maximum speed.odelled in Simulink.
In this research a simulation model in MATLAB/Simulink for the sensorless position/speed estimation and closed-loop control of a Permanent Magnet Synchronous Machine (PMSM) used in the steer-by-wire implementation is proposed. The presence of a sensorless algorithm in the steer-bywire application improves the safety and robustness of the application. The model presented as part of this research is intended to facilitate the design of the various Proportional (P) and Propor-tional Integral (PI) controllers used in practice in both the vector control algorithm of the PMSM and the sensorless observer. The model also allows the prediction of the steer-by-wire performance under different load conditions. Since model based sen-sorless control systems
Figure 1:
Steer-by-wire system.
2 STEER-BY-WIRE SYSTEM 2.1 Steer-by-wire System Overview The steer-by-wire system aims to eliminate some of the mechanical linkages within a traditional steering mechanism while still replicating the steering feel at the driver’s end. The direct mechanical coupling is eliminated and re-placed with two electrical drives with associated position and speed sensors such as encoders as shown in Figure 1. The two electrical drives are required for providing torque feedback at the handwheel (Machine M1) and replicating the position of the handwheel at the steered side (Machine M2).
2.2 Characteristics of Steer-by-Wire The numerous steer-by-wire systems proposed to date have shown considerable advantages over conventional steering arrangements. The absence of the steering column simplifies interior car design and therefore the handwheel can be placed at different parts of the vehicle since it does not have to be necessarily mounted on the dashboard. The removal of the physical link be-tween the handwheel and the steered wheel ar-rangement also results in a lower probability that the
Figure 2:
PMSM modelled in Simulink.
April 2017 ISSUE 56
33
Sensorless Position Control of a PMSM for Steer-by-Wire Applications Continued
impact produced by a frontal crash will cause the steering arrangement to invade the driver’s space; hence reducing the occurrence of driver inju-ries. Since the steer-by-wire system requires a digital signal pro-cessor/microcontroller to control the drive it is also possible to adapt the steering dynamics according to the driver’s comfort and preference. One of the main challenges in by-wire applications in general is that of generating an au-thentic force feedback at the driver inputs. In steer-by-wire research considerable effort has been done in order to provide a torque feedback at the handwheel which has the same physiological effects on the driver [79]. The possi-ble failure of the position / speed sensor has also resulted in researchers having to introduce mechanical systems and sensors to com-ply with automotive safety requirements. 3 SENSORED POSITION CONTROL OF THE PMSM The steer-by-wire application requires the design of two control systems for the PMSMs pre-sent in the system. The machine which provides torque feedback at the handwheel must operate in current control mode while the machine which sets the position at the steered wheel side must operate in position control mode. In this research Rotor Flux Orientated Control of the PMSM is used and consists of cascaded control loops for position, speed and current. The parame-ters of the PMSMs considered for the steer-by-wire application re-quired for the generation the MATLAB/Simulink model are listed in Table I. The open-loop bandwidth of the electrical part of the PMSM model has a bandwidth of 65 Hz while the open-loop bandwidth of the mechanical part has a bandwidth of 0.016 Hz. The bandwidths were determined from the s-domain transfer functions for the respective models. The controllers in the cascaded position control system for the PMSM where tuned as listed in Table II. The gains chosen for the controllers were tuned in such a way that the PMSM exhibits a stable response when in sensorless control mode. The PI controllers in the system have a dual nature; they are used to provide an adequate response with minimum delay while filtering out any noise present due to the sensorless estimates.
Description
Value
Unit
Ld
D frame stator inductance
807.9
mH
Lq
Q frame stator inductance
641.1
mH
R
Stator Resistance
262
J
Moment of Inertia
7.7 × 10
B
Coefficient of Friction
p
Number of Pole Pairs
Table 1:
PMSM model parameters.
34
3 SENSORLESS POSITION CONTROL OF THE PMSM 3.1 Sensorless Position Control Overview The use of model-based sensorless observers is widely reported in literature due to their applicability over a range of speeds; however most of these observers fail at low or zero speed conditions [10]. These methods rely on the estimate of the back-emf, which decreases significantly at low speeds, hence the resulting estimates are susceptible to sensor noise and variation in machine parameters. Since in steer-by-wire, position control at low speed is required, model-based observers are not suitable. Hence sensorless algorithms which are based on tracking machine saliencies must be used in order to obtain accurate rotor position/speed estimates. Tracking of saliencies is typically done through the injection of additional signals during operation of the electrical machine. The methods for signal injection are subdivided into transient and continuous injection methods. In this research a continuous high frequency (HF) signal injection method is used. HF-based methods in literature have been shown to obtain a rotor position estimate by superimposing a continuous high frequency carrier on the voltage fed onto the PMSM for the purpose of saliency tracking [11, 12]. 3.2 High Frequency Injection Saliency Tracking In the MATLAB/Simulink model presented in this article sensorless control is based on injecting a high frequency component in the stationary αβ-frame. These components are added to the reference fundamental frequency stator voltages in the synchronous dq-frame transformed into the αβ stationary as shown in Figure 3. The three phase currents ia, ib and ic contain components of both the fundamental frequency and the injected HF. For the purposes of saliency tracking the low frequency current component is filtered out with a fourth order band-pass filter and transformed to the αβ frame. For an injection frequency ωi of 2 kHz a fourth order band-pass filter with double poles at 1 kHz and 3.3 kHz respectively was designed and included in the model. The output of this filter after transformation is denoted as iiαβ and is the isolated current component due to the HF injection on the stator windings. The heterodyning approach is then used on iαβ to obtain an error term which can be used in a phasedlocked loop (PLL) to obtain the speed and position estimates as shown in Figure 4.
Loop
Kp
Ki
Closed-Loop Bandwidth [rad/s]
Damping Ratio
kg m2
Current
0.55
365.13
753
0.84
7.5 × 10-4
Nms
Speed
1.92
113.14
82
0.707
6
-
Position
10
0
10
1
mΩ -3
Table 2:
Controller gains.
4 CLOSED-LOOP POSITION/SPEED TRACKING SIMULATION RESULTS The main aim of the PMSMs in the steer-by-wire system is to track position references at low speed. The Simulink/ MATLAB model developed as part of this research has inputs which allow for testing of this tracking capability for different position references and different load con-ditions.
Table 3:
High Frequency Injection in the αβ stationary frame.
In the simulation results presented in this arti-cle a full-load of 5 Nm is applied at the output of the PMSM. A change in reference posi-tion from 0.1 rad to 1.1 rad is applied at 0.5 s and filtered through a first order low-pass filter with a bandwidth of 25Hz. The test was carried out while operating in sensorless control. The actual/estimated shaft speed is shown in Figure 5. The actual/estimated shaft position is shown in Figure 6. The three phase stators currents and the dq frame stator currents are shown in Figures 7 – 8. 5 CONCLUSIONS In the simulation results presented the track-ing of the rotor position/speed for a change in ref-erence position at full load conditions was shown. The sensorless observer presented and simulated in this article tracks both speed and position with minimum error. During the operation shown in Section 4 the system was operating in sensorless control, that is the estimated speed and position where being both observed and used for control purposes. In this mode the error in the estimates was negligible for the purposes of vector control. The change in the position reference given in simulation is of 1 rad and reaches steady state in approximately 39 ms. This translates into a steering rate of change of 1458 degrees/s which is well above the average human steering capability observed as part of this same research.
Table 4:
PLL rotor position/speed estimator.
Table 5:
Table 7:
Plot of Actual/Estimated PMSM Shaft Speed.
Plot of PMSM Stator Currents.
Table 6:
Table 8:
Plot of Actual/Estimated PMSM Shaft Position.
Plot of PMSM Stator DQ Frame Currents.
April 2017 ISSUE 56
35
Sensorless Position Control of a PMSM for Steer-by-Wire Applications Continued
Hence the model presented and designed in this paper shows the validity of using a HF injection based saliency detection method to estimate the speed and position in sensorless steer-by-wire PMSMs. The inclusion of such algorithms could result in improved functional safety in the steerby-wire application by providing an alternative means of measurement for shaft positon/speed. 6 REFERENCES
[1] E. A. Bretz, "By-wire cars turn the corner," Spec-trum, IEEE, vol. 38, pp. 68-73, 2001. [2] M. Bertoluzzo, G. Buja, and R. Menis, "Control schemes for steer-by-wire systems," Industrial Elec-tronics Magazine, IEEE, vol. 1, pp. 20-27, 2007. [3] M. Bertoluzzo, P. Bolognesi, O. Bruno, G. Buja, A. Landi, and A. Zuccollo, "Drive-by-wire systems for ground vehicles," in Industrial Electronics, 2004 IEEE International Symposium on, 2004, pp. 711-716 vol. 1. [4] P. L. Jansen and R. D. Lorenz, "Transducerless position and velocity estimation in induction and salient AC machines," Industry Applications, IEEE Transactions on, vol. 31, pp. 240-247, 1995. [5] S. Zhang, X. Jin, J. Junak, D. Fiederling, G. Saw-czuk, M. Koch, et al., "Permanent magnet technol-ogy for electric motors in automotive applications," in Electric Drives Production Conference (EDPC), 2012 2nd International, 2012, pp. 1-11. [6] S. Bolognani, M. Tomasini, and M. Zigliotto, "Control Design of a Steer-byWire System with High Performance PM Motor Drives," in Power Elec-tronics Specialists Conference, 2005. PESC '05. IEEE 36th, 2005, pp. 1839-1844. [7] A. Baviskar, J. R. Wagner, D. M. Dawson, D. Braganza, and P. Setlur, "An Adjustable Steer-by-Wire Haptic-Interface Tracking Controller for Ground Vehicles," Vehicular Technology, IEEE Transactions on, vol. 58, pp. 546-554, 2009. [8] S. M. H. Fahami, H. Zamzuri, and S. A. Mazlan, "Development of Estimation Force Feedback Torque Control Algorithm for Driver Steering Feel in Vehicle Steer by Wire System: Hardware in the Loop," International Journal of Vehicular Technology, vol. 2015, 2015. [9] Y. Yamaguchi and T. Murakami, "Adaptive Con-trol for Virtual Steering Characteristics on Electric Vehicle Using Steer-by-Wire System," Industrial Electronics, IEEE Transactions on, vol. 56, pp. 1585-1594, 2009. [10] C. S. Staines, C. Caruana, and R. Raute, "A Review of Saliency-based Sensorless Control Meth-ods for Alternating Current Machines," IEEJ Journal of Industry Applications, vol. 3, pp. 86-96, 2014. [11] J.-I. Ha and S.-K. Sul, "Sensorless field-orientation control of an induction machine by high-frequency signal injection," Industry Applications, IEEE Transactions on, vol. 35, pp. 45-51, 1999. [12] J. Cilia, G. M. Asher, K. J. Bradley, and M. Sumner, "Sensorless position detection for vector-controlled induction motor drives using an asymmetric outer-section cage," Industry Applications, IEEE Transactions on, vol. 33, pp. 1162-1169, 1997.
Mr Kris
Scicluna
Kris Scicluna graduated with a B. Eng. Hons. in 2011 and with an M.Sc. Eng. in 2013 from the University of Malta. Currently he is following a Ph.D research project with the Department of Ind. Elec. Power Conversion at the Faculty of Engineering of the University of Malta entitled “Sensorless Control in Steer-by-Wire Applications�. He is also a lecturer at the Institute for Engineering and Transport, Electrical and Electronics, MCAST. His research interests include power electronics, electrical drives and renewable energy.
Prof. Dr Ing. Cyril
Spiteri Staines
Cyril Spiteri Staines graduated with a First Class B.Eng. degree from the University of Malta in 1995 and with a Ph.D. degree from the University of Nottingham in 1998. During 2003-2004 Prof. Spiteri Staines was a visiting scholar at the School of Electrical and Electronic Engineering at Nottingham. He is a Full Professor with the Department of Ind. Elec. Power Conversion of the Faculty of Engineering of the University of Malta. His current research interests are related to design and control of Electrical Machines, Power Electronic Converters, Microgrids, Energy Efficiency and Renewable Energy Sources (RES).
Dr. Ing. Reiko
Raute
Dr. Ing. Reiko Raute is a lecturer with the Department of Ind. Elec. Power Conversion of the University of Malta. His current research interests include power electronics and electric drives.
36