Hitachi to Launch "Optimal Operation Guidance System" by Year-End, Supporting On-Site Production Operations with Physical AI Incorporating Skilled Operator Skills

Image of the "Optimal Operation Guidance System"

Hitachi, Ltd. (hereinafter referred to as Hitachi) will launch the "Optimal Operation*2 Guidance System" (hereinafter referred to as this system) by the end of the year, which supports on-site production operations for chemical manufacturing plants with Physical AI*1. In process manufacturing industries such as chemicals, batch production*3 is essential for increasing added value. In this batch production, the internal state of equipment changes moment by moment, and complex reactions require control, thus production has strongly depended on the experience and intuition of operators. As a result, not only have variations in quality and productivity occurred, but the personal burden of operators needing to predict internal states based on instrument readings has been a major challenge on-site. To solve these challenges, this system, one of the HMAX Industry lineup, reproduces and analyzes the batch production process for manufacturing high-value-added chemicals in small quantities and with high variety in cyberspace, and proposes optimal control methods for DCS*4 and PLC*5.

Specifically, for the plant's reaction equipment*6, which is important in the production process, the system visualizes the state of materials reacting inside the equipment, predicts the impact of specific controls on the materials, and provides operational guidance such as temperature, pressure, and flow rate to the operator. By utilizing this system, operations that are less dependent on the skill and experience of operators become possible, contributing to the stabilization of product quality and the improvement of overall production efficiency. Operators will also be able to perform operations based on predictions of future changes.

This system was realized by combining technology that incorporates domain knowledge of the industrial field into AI with technology that uses reinforcement learning*7 with past operational data obtained from digitized assets. This is the first step towards realizing automated plant operation, which further evolves OT (control and operational technology) provided by Hitachi to manufacturing customers through DCS and PLC with Physical AI.

Hitachi's Connective Industries (CI) Sector, Industrial Solutions Business Unit, is focusing on "HMAX Industry," a next-generation solution suite that combines domain knowledge and advanced AI with the rich installed base of product data (digitized assets). Aiming to be a leading company in Physical AI, we strive to realize a prosperous society by maximizing customer lifetime value and transforming industries globally through the provision of "Industrial Solutions" centered on these.

*1 Physical AI: AI with the ability to recognize and understand the real world, make autonomous judgments, and deploy them into actual actions.

*2 Optimal Operation: Operation that can efficiently reach target values (KPIs) while complying with constraints, with respect to the target values (KPIs) and constraints defined in production.

*3 Batch Production: A production method suitable for small-lot, high-mix production, where the processes of raw material input, manufacturing, and product removal are carried out one batch at a time. Used in contrast to continuous production.

*4 DCS (Distributed Control System): Distributed Control System

*5 PLC (Programmable Logic Controller): A device for controlling the operation of machinery and equipment with a program.

*6 Reaction Equipment: Plant equipment used to manufacture products from raw materials through reaction.

*7 Reinforcement Learning: A technology that allows Physical AI to learn rules and make optimal judgments based on a large amount of input/output data.

Features of this System

This system was realized by combining the following two models*8:

1. Process Model Fusing Domain Knowledge and AI

Hitachi has developed technology to incorporate domain knowledge of R&D and control technology into AI. This is expected to visualize the internal processing of AI, which was previously opaque in reaching predictions, and support operator judgment.

In predicting the internal state of reaction equipment, the energy and material inflows and outflows are calculated using theoretical formulas based on Hitachi's chemical and chemical engineering expertise. For parts that are difficult to formalize, such as equipment degradation over time, AI is used for prediction. This has made it possible to visualize the internal calculation results based on theoretical formulas. This technology allows for the incorporation of on-site knowledge into AI, and it has been confirmed that accurate predictions can be made with less data than with conventional AI technologies built by Hitachi.

Furthermore, typically, such theoretical formulas need to be reviewed for each manufacturing process, which takes time to build models. However, Hitachi has established a mechanism for generally applying theoretical formulas. This is expected to shorten model building time and contribute to improving lead time until delivery.

2. Control Model Capable of Efficiently Proposing Optimal Control Methods

Through Hitachi's proprietary technology of reinforcement learning on operational data from past batch production, the system proposes equipment setting candidates to the operator based on conditions that resulted in good operational outcomes, such as fewer error occurrences, high product quality, and short operating times.

Generally, in batch production, the operational state changes over time. Therefore, it is costly and time-consuming to have AI learn and predict the optimal operation at each moment. Hitachi has therefore developed technology for efficient learning and prediction. This technology defines states using its proprietary clustering technology and performs reinforcement learning on the time-series changes of these defined states, enabling learning and prediction in a relatively short period compared to reinforcement learning methods that Hitachi has previously considered.

*8 The technology related to this system is patented. The patent status is as of the date of this press release and may change in the future.

Background

Process industries such as chemicals are undergoing rapid technological innovation and globalization. To compete with inexpensive products from overseas, batch production of small-lot, high-mix production of high-value-added products is essential. Batch production involves independent processes, and manual operations such as changing operating conditions and inputting/exchanging raw materials are prone to occur, requiring skilled operators. Due to variations in product quality, manufacturing time, and productivity depending on the operator's skill level, there is a demand to move away from manual operations and improve productivity.

Related Site

Optimal Operation Guidance System

https://www.hitachi.co.jp/products/infrastructure/product_site/ope-guide/index.html

Trademark Notice

Company names and product names mentioned are trademarks or registered trademarks of their respective companies.

About Hitachi, Ltd.

Hitachi aims to be a global leader that continues to innovate social infrastructure digitally through its Social Innovation Business (SIB) utilizing IT, OT (control and operational technology), and products, and contributes to the realization of a Harmonized Society where environment, well-being, and economic growth are in balance. With a business structure comprising four sectors: Digital Systems & Services, Energy, Mobility, and Connective Industries, as well as a strategic SIB Business Unit that creates new growth businesses, Hitachi operates globally and solves customer and societal challenges by creating value from data with Lumada*9 at its core. For fiscal year 2025 (ending March 2026), revenue is projected to be 10,586.7 billion yen. As of the end of March 2026, Hitachi has 606 consolidated subsidiaries and approximately 290,000 employees worldwide. For more information, please visit www.hitachi.com/ja-jp/.

*9 Lumada: The collective name for Hitachi's advanced digital solutions, services, and technologies for creating value from customer data and accelerating digital innovation.

Contact Information

Hitachi, Ltd.

Connective Industries

Inquiries about Products and Solutions