Use of artificial intelligence (AI) in supply chain management and study of industry use cases of AI in factory capacity optimisation

Abstract

Artificial Intelligence (AI) based technologies are currently disrupting all the industries due to the development of ¢®jsors and advanced processors capable of collecting and analyzing vast amounts of data. Al is playing a pivotal role in increasing the visibility and flexibilities of the supply chain. Al is applicable to every aspect of the supply chain, from demand estimation to the last mile. This study aims to study implemented AI/ML methods in various industries and subsequently chart out the state of Al in different industries and the functions. The AI tools like feed forward networks and tree-based ensemble learning are widely used in SCM. After studying the role of AI in SCM, we delved into specific use cases of Al applications in Manufacturing operations, mainly quality inspection, maintenance, and process optimization. Most of the industry use cases we found centered on the application of machine learning (ML), Deep Learning, and Autonomous Objects. To understand the applicability of AI/ML in quality inspection, we analyzed solutions implemented in chip (Micron) and beverage (Carlberg) industries. Micron implemented computer vision combined with neural networks and deep learning algorithms to detect and categorize the anomalies in chips. While Carlsberg implemented a beer fingerprint project using sensors for collecting flavor and aroma data to map the beer flavors with the ingredient composition. We analyzed solutions implemented in the Automobile industry (General Motors), Gear Manufacturing (ZF Friedrichshafen AG (ZF), and aerospace industry (Airbus). GM implemented a computer vision method for early detection of failure of its assemf§Jy robots to save on production downtime. ZF, a pioneer in gear manufacturing, deployed support vector machine (SVM), random forest, and neural network techniques to train linear regression models for detecting the failure of honing tools. Airbus implemented neural network and physics-based models for tracking the health of its starters which play an important role during the start-up phase by supplying power to auxiliary power units. We draw important insights by studying process optimization solutions implemented by 3B Fiberglass and FANUC. 3B fiberglass used the sensors to collect the process parameters combined with computer vision for early detection and pinpoint the root cause of fiber breakage by comparing the images with the database. FANUC deployed a deep reinforcement learning network to make the robots learn by themselves without requiring manual programming and drastically reduced the training time. Going further, we have narrowed our scope to factory capacity optimization. We will analyze the solutions implemented around digital twin for process development & control and lot lineage for tracking material sources for quality control.