An intelligent self-healing method based on reinforcement learning and control performance monitoring

Abstract

The invention discloses an intelligent self-healing method based on reinforcement learning and control performance monitoring. The present invention establishes a dynamic model for the manufacturing process to be studied, establishes a corresponding reward function and an objective function, and combines a neural network to respectively construct a value function evaluation network and a set value action network, which solves the problem that the set value is difficult to adjust continuously , calculate the dynamic monitoring index on-line, monitor the control performance of the controller, and provide the variable period control basis for the self-healing controller. This method overcomes the high cost problem of re-establishing an accurate process dynamic model, and effectively solves the problem of online quality control of end products. By changing the control cycle, problems such as instability and overshoot caused by frequent controller adjustments can be avoided, ensuring The quality of the end products in the manufacturing process helps to improve product quality stability and production efficiency, thereby improving the economic benefits of industrial production.

Description

technical field [0001] The invention belongs to the field of on-line control of industrial intelligent manufacturing process quality, in particular to an intelligent self-healing method based on reinforcement learning and control performance monitoring. Background technique [0002] In recent years, with the rise of artificial intelligence and industrial big data, more and more traditional manufacturing processes have begun to transform and upgrade to intelligent manufacturing processes. For the intelligent manufacturing process, intelligent optimization operation is the core of the control problem. However, changes in production conditions and other unknown factors often lead to complex and unknown changes in the production process, resulting in a large number of unqualified and defective products, causing huge industrial losses, and even affecting the quality of the production process. Safe operation, leading to various production accidents. For some manufacturing proces...

AI Technical Summary

Problems solved by technology

However, at present, many industrial processes still use open-loop control or hierarchical feedback control. On the one hand, there is a lack of breakthroughs in new control methods, and on the other hand, the control performance is limited due to economic factors.

That is, traditional control methods cannot meet the control performance requirements, and the control problems of complex industrial processes are waiting for breakthroughs in new technologies

Images