The Autonomous Production System, which utilizes two proprietary AI applications—PCM (Plant Capability Maximizer) and APS (Advanced Prediction System)—does not optimize plant operations automatically simply by being implemented. This is because gaps can arise between optimizations derived from the know-how of seasoned operators-programmed into the AI- and actual performance data from the production floor. It is important to assiduously verify these gaps and uncover new know-how and insights that had previously gone unnoticed. This process leads to the clarification of weak points in our manufacturing operations as well as the reaction logic based on chemical engineering principles that has not yet been fully established as technology.
The PCM can predict quality status in real time, which has made it possible to aggressively pursue operating conditions that meet both quality and cost needs. In manufacturing at chemical plants, raw material quality is not always uniform, making the control of raw materials difficult. For example, pulp, the raw material for cellulose acetate, is a natural product and is easily affected by temperature and humidity, making it a challenging material to handle. Previously, the sentiment was that ensuring consistent quality was the priority, even if the costs were somewhat higher, in order to achieve stable production.
However, the PCM not only allows quality status to be understood at a glance but also predicts quality when conditions are changed, removing any resistance to adjusting those conditions. Previously, there was no perceived need to take risks by adjusting conditions as long as production was stable, whereas it is now possible to make decisions to improve each indicator even if just slightly.
A major effect of the APS (Advanced Prediction System) is a significant reduction in response time when an abnormality is detected. For example, when the performance of some equipment in chemical plants dropped, operators previously had to rush to the floor to check the situation, and time was needed after that to consider countermeasures. In chemical plants, halting equipment means halting production itself, so operators search for ways to recover by adjusting other settings without halting equipment. However, because decisions had to be made based on a variety of complex and interrelated conditions at that time, the consideration of countermeasures required a lot of time.
The APS, on the other hand, continuously monitors the conditions of equipment and immediately indicates the necessary recovery actions if a drop in equipment performance is about to occur. Tasks that were previously driven by reactive responses can now be shifted to proactive work.
In the Autonomous Production System, the AI logic has been developed with the aim of preventing it from becoming a black box. One of the key elements that made this possible is Daicel’s proprietary logic tree, which was created based on a Comprehensive Operability Study. In this proprietary logic tree, causal relationships, from cause to effect, are organized into a tree structure for abnormalities occurring in the plant. In other words, whereas we previously looked at one step before and after, this has now been extended to looking at five steps before and five steps after so the cause-and-effect relationships can be better understood.
To optimize plant operations or detect operational abnormalities, multiple factors must be taken into consideration. For operators, it can sometimes be difficult to understand why a certain result occurred. By visualizing the phenomena that occur based on Daicel’s proprietary logic tree, it becomes possible to understand the flow of cause-and-effect relationships, almost as if looking inside the mind of a seasoned operator.
Furthermore, by promptly having AI learn and utilize new know-how, machines evolve while people grow by achieving even more advanced manufacturing. The Autonomous Production System is a system that creates a continuous cycle in which people and machines grow together.
Objectively viewing and understanding the principles (connections of cause-and-effect relationships) presented by AI without being bound by conventional wisdom provides the push to execute ideas like "Wouldn't it be better if we did it like this?" or " Wouldn't it be possible to try this approach next time?" As an example, it has led to major improvements, including the optimization of raw material usage rates and other factors in the manufacturing of cellulose acetate. Through implementation of the Autonomous Production System, we believe that the work of operators will change even further in the near future.
At present, monitoring and detection of abnormalities in plant operation are performed by people; however, AI will also be applied to the know-how involved in such abnormality detection. By further increasing the range of tasks that can be entrusted to machines, operators should be able to devote more time to creative work that only people can perform. In a market where needs are constantly changing, we are aiming for a production floor that thinks for itself and collaborates with sales, procurement, purchasing, and logistics departments, as well as with customers, to respond to change and create added value unique to Daicel. By increasing opportunities for operators to actively test their own ideas through the Autonomous Production System, Daicel’s manufacturing capabilities will become even stronger.
