By Chetan Gupta, Ph.D.
R&D Division, Hitachi America, Ltd.
I am thrilled to be writing the inaugural article welcoming you to our new blogsite focusing on "Industrial AI". For more than a century, Hitachi has been a leader in the industrial domain with products, services and solutions in numerous verticals such as railways, automotive, power and energy systems, elevators and escalators, mining equipment, healthcare, etc. Besides being an industrial giant, we have also excelled in the information technology domain. This rich heritage in operational technology (OT) and informational technology (IT) places us in a unique position to build solutions that lie at the intersection of these two fields.
A key component of many of these solutions will be Industrial AI. Industrial AI is concerned with the application of Artificial Intelligence (AI), Machine Learning (ML) and related technologies towards addressing real world challenges in industrial and societal domains. These challenges can be categorized into the horizontal areas of product and service quality, operations optimization, maintenance and repair, safety, supply chain and logistics and end-to-end optimization. The data obtained from real world physical systems are often heterogenous with the more prominent data types being sensor data, event data, images and text, and unlike many of today’s examples of successful application of AI, can often be limited in size. The techniques we use to address the challenges given the data vary from traditional statistics and operations research-based methods to some of latest cutting-edge work in deep networks.
Industrial AI will have far reaching impact on the world that we live in. For example, in AI-driven fleet management data coming from the fleet are analyzed in real-time to predict service interruptions and recommend the necessary repairs, in smart manufacturing defective products are detected and discarded early in the production line avoiding costly production processes, and in smart operations real-time data captured from operators and operations is utilized to recommend an optimal operating envelope for performing the task at hand.
Through this blog, my colleagues and I in the Hitachi global research organization, hope to share not only our latest research results but also real-world applications of Industrial AI. For example, in the first few posts we will share insights on using a deep network to detect anomalous sounds for industrial inspections (similar to what is done manually today), and using deep reinforcement learning to detect performance degradation in equipment.
Hopefully, the reader can take away from this blog new lessons and insights about this deeply fascinating and profoundly important area. Happy reading!
