An Overview of Data Science in Manufacturing Data science has greatly increased various industrial applications over the last few years. Data science is now used in health care, customer service, government, cybersecurity, mechanical, aerospace, and other industrial applications. Among these, manufacturing has grown in importance to achieve the simple goal of Just-in-Time delivery (JIT). Manufacturing has gone through four major industrial revolutions in the last 100 years. We are currently in the fourth Industrial Revolution, where data from machines, the environment, and products is harvested to get closer to Just-in-Time's simple goal: "making the right products in the right quantities at the right time." The obvious solution is to lower production costs so that goods may be sold at lower prices. In this article, I will attempt to answer some of the most frequently asked data science in manufacturing questions.
What is the impact of data science on the manufacturing industry? There are numerous applications of data science in manufacturing. Predictive maintenance, predictive quality, safety analytics, warranty analytics, plant facility monitoring, computer vision, sales forecasting, KPI forecasting, and many more are just a few examples.
● Maintenance Prediction: Manufacturing machine breakdown is extremely costly. The single largest contributor to manufacturing overhead costs is unplanned downtime. Over the last three years, unplanned downtime has cost businesses an average of $2 million. In 2014, the average cost of downtime per hour was $164,000. By 2016, that figure had risen by 59% to $260,000 per hour. This has resulted in adopting technologies such as condition-based monitoring and predictive maintenance. Continuous tracking of sensor data from machines is performed to detect anomalies (using models such as PCA-T2, one-class SVM, autoencoders, and logistic regression), diagnose failure modes (using classification models such as SVM, random forest, decision trees, and neural networks), and predict the time to failure (TTF) (using a combination of techniques such as survival analysis, lagging, curve fitting and regression models)
● Vision in computers: Traditional computer vision systems measure parts for tolerance to determine whether or not they are acceptable. Inspecting the claims for defects such as scuff marks, scratches, and dents is also critical. Humans have traditionally been used to scan for such flaws. Today, AI technologies such as CNN, RCNN, and Fast RCNNs have proven to be more accurate than
their human counterparts while checking in a fraction of the time. As a result, the cost of the products is significantly reduced.
● Forecasting sales Predicting future trends has always aided in resource optimization for profitability. This has been observed in various industries, including manufacturing, airlines, and tourism. Knowing the manufacturing volumes ahead of time aids in optimizing resources such as supply chain, machine-product balancing, and workforce. Today, techniques ranging from linear regression models, ARIMA, and lagging to more complex models such as LSTM are used to optimize resources. Refer to the industry-accredited data science course for more information.
● Quality prediction: The quality of the products produced by the machines is predictable. As shown in Figure 2, statistical process control techniques are the most common tools used on the manufacturing floor to determine whether a process is under control or out of control. A good trend line could be obtained using statistical techniques such as linear regression on time and product quality. This line is then extrapolated to answer questions like "How long before we start making bad parts?"
What tools do data scientists in the manufacturing industry use? A data scientist in manufacturing employs a variety of tools throughout the project lifecycle. As an example: Notebooks (R markdown and Jupyter), GIT, and PowerPoint were used for the feasibility study. "Yes! You read that correctly. PowerPoint is still essential in any organization. BI tools are attempting to usurp them. In my experience with a half-dozen BI tools, PowerPoint remains the best tool for storytelling." If you aspire to become a data scientist or analyst, take up a data science course in Mumbai and become job-ready in less than 6 months.