Data science in manufacturing
There has been a trend in the past years that data science has penetrated various industrial applications. Data science is applied in health care, cybersecurity, governments, aerospace, among other industrial applications. In manufacturing industry, data science is solving the roadblocks to achieve the Just-in-Time (JIT) manufacturing. JIT is a production model in which items are created to meet demand, not created in surplus thus helping manufacturing companies to reduce costs, increase efficiency, and speed up product delivery. (Robert Sheldon)
Over the last 200 years we have experienced four industrial revolutions and the speed of current breakthroughs has no historical precedent. We are witnessing the Fourth Industrial revolution which is evolving at an exponential rather than a linear pace. (Schwab, 2016) Current Industrial revolution has given birth to Industry 4.0 (or modern manufacturing) which is revolutionizing the way companies manufacture, distribute and improve their products. (What is Industry 4.0?) Industry 4.0 is enabling industries to get closer to Just-in-Time as data from machines, products and environment is being collected and analyzed to achieve precise results.
Data science is envisioned to revolutionize manufacturing industry dramatically. Following are the data science use cases in manufacturing that have been incorporated by the manufacturers.
1. Predictive analytics
Predictive analytics is the analysis of present data to forecast and avoid possible future problems in advance. Manufacturers seek to solve issue like downtime, logistics, over production, inventory & idle time. Predictive analysis provides solutions to solve these problems.
Fault prediction and preventive maintenance
There are 2 types of preventive maintenance: time-based and usage based. Preventive maintenance helps in planning as it provides prediction concerning possible future breakdown of equipment. To mitigate the breakdown, the manufacturer may plan a shutdown for repairing or plan a break. Such breaks are usually made to avoid considerable delays and failures which can be caused by major breakdowns.
Machine breakdown results in unplanned downtime and is the single largest contributor to manufacturing overhead costs. The average cost of downtime is significant and each minute costs an average of $9,000 bringing the downtime cost per hour to over $500,000. (Ponemon Institute, 2016) Predictive maintenance and condition-based monitoring can help to mitigate machine breakdown and reduce downtime costs.
Sensor data from machines are monitored continuously 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), predict the time to failure (TTF) (using combination of techniques such as survival analysis, lagging, curve fitting and regression models) and optimal maintenance time prediction (using operations research techniques). (Nagdev Amruthnath, 2018) and (Amruthnath & Gupta, 2018)
2. Data science in supply chain
A supply chain has several activities that are necessary for producing and delivering products or services to customers. Due to number of actors in a supply chain, they are complex, unpredictable and entails risk. Manufacturing companies use raw materials to process into finished goods. Data analytics in materials management can help to optimize processes such as sourcing, quantity, storage, safety, and quality check. It also measures the finished products quality standard and analyzes the impact of raw materials on the manufacturing process. Analytics also help companies to predict potential delays and calculate probabilities of the problematic issues. It helps companies to develop contingency plans and identify back up suppliers. (Data Scientists in Supply Chain Management, 2022)
3. Computer Vision Applications
Computer vision applications and AI technologies are used in manufacturing during the stage of quality control. These technologies efficiently enabled object identification, object detection and classification. Usually, quality control monitoring was performed by humans for quality control for defects such as scuff marks, scratched and dents. AI technologies such as CNN, RCNN, and Fast RCNN’s have proved their accuracy as compared to humans for quality inspection & control thereby reducing the costs. (Tian Weng, 2018)
4. Demand forecasting and inventory management
Demand forecasting and inventory management are important components of data science because they involve the use of statistical and machine learning techniques to analyze data and make predictions or decisions about future demand and inventory levels.
In demand forecasting, data scientists can use historical sales data, market trends, and other relevant information to predict future demand for a product or service. Demand forecasting uses the data from the supply chain and has strong relations with inventory management. This interrelation exists because the supply chain data is used for demand forecasting. This can help a company better plan for production, marketing, and other activities to meet customer demand.
In inventory management, data scientists can use demand forecasting and other data sources to optimize inventory levels and improve efficiency. For example, they might use data on lead times, safety stock levels, and demand patterns to determine how much inventory to keep on hand and when to restock.
Overall, demand forecasting and inventory management can help data scientists and organizations make more informed decisions about resource allocation, production, and other business operations, ultimately leading to improved efficiency and profitability.
5. Price optimization
Price optimization is the process of using data and statistical techniques to determine the optimal price for a product or service. In data science, price optimization can be achieved through a variety of methods, including the use of machine learning algorithms and statistical modelling.
One common approach to price optimization is to build a machine learning model that considers a variety of factors that might influence price, such as the demand for a product, the cost of production, and the prices of competitors. The model can then be trained on historical data to learn the relationships between these factors and the optimal price.
Another approach to price optimization is to use statistical modelling techniques, such as regression analysis, to identify patterns in the data and build a model that can predict the optimal price based on these patterns.
In both cases, the goal is to use data science to identify the factors that most influence price and to develop a model that can accurately predict the optimal price based on these factors. This can help manufacturers make more informed pricing decisions and maximize their profits.
Data science in manufacturing is relatively new as compared to other disciplines. In manufacturing projects (capital investments), ROI is realized in 5-7 years. Successfully deployed data science projects have ROI in less than a year. Data science is one of the tools that manufacturing industries are currently using to achieve their JIT goal.
Companies adopting data science currently experience an ROI of 4.4x and expect that figure to 6.7x in next two to three years. There is a difference between in ROI between the early and late adopters of data science. Early adopters see ROI of 5.8x while late adopters currently see only 3.8x. Any increase in ROI is beneficial but the difference of 2x makes a clear case for getting models into production as soon as possible. (How Innovators Use Advanced Analytics, AI, and Machine Learning to Increase ROI & Competitiveness, 2021)
Amruthnath, N., & Gupta, T. (2018). A research study on unsupervised machine learning algorithms for early fault detection in predictive maintenance. 5th International Conference on Industrial Engineering and Applications (ICIEA). Singapore: IEEE.
Data Scientists in Supply Chain Management. (2022, June 27). Retrieved from Industry Today: https://industrytoday.com/data-scientists-in-supply-chain-management/
How Innovators Use Advanced Analytics, AI, and Machine Learning to Increase ROI & Competitiveness. 2021. Rapidminer. Retrieved from https://rapidminer.com/downloads/data-driven-transformation/
Nagdev Amruthnath, T. G. (2018, March). Fault class prediction in unsupervised learning using model-based clustering approach. International Conference on Information and Computer Technologies (ICICT). Dekalb, Illinois: IEEE. Retrieved from https://ieeexplore.ieee.org/document/8356831
Ponemon Institute. (2016). Cost of Data Center Outages. Ponemon Institute Research Report. Retrieved from https://www.vertiv.com/globalassets/documents/reports/2016-cost-of-data-center-outages-11-11_51190_1.pdf
Robert Sheldon. (n.d.). just-in-time manufacturing (JIT manufacturing). Retrieved November 10, 2022, from TechTarget: https://www.techtarget.com/whatis/definition/just-in-time-manufacturing-JIT-manufacturing
Schwab, K. (2016, January 14). The Fourth Industrial Revolution: what it means, how to respond. Retrieved November 10, 2022, from World Economic Forum: https://www.weforum.org/agenda/2016/01/the-fourth-industrial-revolution-what-it-means-and-how-to-respond/
Tian Weng, Y. C. (2018). A fast and robust convolutional neural network-based defect detection model in product quality control. The International Journal of Advanced Manufacturing Technology, 94.
What is Industry 4.0? (n.d.). Retrieved November 10, 2022, from IBM: https://www.ibm.com/topics/industry-4-0