Search

CN-122018476-A - Industrial control parameter optimization system and method based on neural symbol double-flow architecture

CN122018476ACN 122018476 ACN122018476 ACN 122018476ACN-122018476-A

Abstract

The invention provides an industrial control parameter optimization system and method based on a neural symbol double-flow architecture, which belong to the technical field of industrial intelligent control and automation. The method and the device realize time scale decoupling of AI cognitive calculation and bottom layer mechanical control, solve the problems of system resource competition and time sequence breakdown caused by the operation of the two same time axes, cut off a direct drive link of a probabilistic AI model to a bottom layer physical actuator, eliminate override control risks, remodel a bottom layer safety defense line of industrial control, construct a low-cost, low-risk and non-invasive intelligent upgrading framework of an adaptive stock old PLC/DCS system, and avoid high implementation risks of modifying bottom layer core control logic.

Inventors

  • LIU JIAYAN
  • LI JINGSI

Assignees

  • 成都思为交互科技有限公司

Dates

Publication Date
20260512
Application Date
20260415

Claims (10)

  1. 1. The industrial control parameter optimization system based on the neural symbol double-flow architecture is characterized by comprising a slow-flow cognitive subsystem, a fast-flow control subsystem and an intermediate data interaction registering module; The slow-flow cognitive subsystem is deployed in edge computing equipment of an industrial field, is internally provided with a small language model subjected to fine adjustment in the industrial vertical field, is used for acquiring multi-mode working condition data of the industrial field in a bypass monitoring mode, and is combined with built-in historical work order data and process expert experience data to execute global optimizing calculation of process parameters with minimum comprehensive production cost as a target so as to generate a process parameter suggestion value; The intermediate data interaction registering module is arranged in an accessible storage area of the fast flow control subsystem and used as a unidirectional communication isolation interface between the slow flow cognition subsystem and the fast flow control subsystem, and is used for receiving and storing a technological parameter proposal value written in the slow flow cognition subsystem, and the writing action of the slow flow cognition subsystem to the intermediate data interaction registering module does not trigger hardware interruption of the fast flow control subsystem; The quick control subsystem is deployed in industrial controller hardware which is closely attached to an industrial field physical execution mechanism, is operated based on an embedded real-time operating system, has the operation characteristic of hard real-time fixed-period cyclic scanning, is endowed with unique physical execution mechanism control right in the system, is forbidden to respond to an external hardware interrupt request and a driving instruction directly issued by the external system, actively pulls a currently stored technological parameter recommended value from an intermediate data interaction registering module in each fixed scanning period, completes deterministic safety verification of the pulled technological parameter recommended value in the same scanning period, processes the technological parameter recommended value into a final control target value after verification is passed, substitutes the final control target value into a built-in closed-loop control algorithm to generate a driving signal, transmits the driving signal to the corresponding physical execution mechanism to complete control action, maintains the safe control target value operation of the previous scanning period when verification is failed, and triggers an abnormal alarm.
  2. 2. The industrial control parameter optimization system based on the neural symbol double-flow architecture according to claim 1, wherein the multi-mode working condition data acquired by the slow-flow cognitive subsystem comprises time sequence working condition data acquired by an industrial field sensor, technical specification text data, equipment maintenance record data and experience record data of field operators, and when the slow-flow cognitive subsystem executes global optimizing calculation of the technological parameters, preset technological quality constraint conditions are synchronously met.
  3. 3. The industrial control parameter optimization system based on the neural symbol double-flow architecture according to claim 1, wherein the intermediate data interaction registering module adopts a nonvolatile data storage area inside an industrial controller or an independent double-port RAM module mounted on an industrial real-time bus, and the intermediate data interaction registering module only opens one-way write-in permission of a designated storage address to a slow-flow cognitive subsystem and only opens read permission of a corresponding storage address to a fast-flow control subsystem.
  4. 4. The industrial control parameter optimization system based on the neural symbol double-flow architecture according to claim 1, wherein the industrial controller adopted by the fast flow control subsystem comprises a programmable logic controller, a distributed control system controller or an independent motion controller, the fixed cycle scanning period of the fast flow control subsystem is 1-20 milliseconds, the time deviation of the scanning period is controlled at microsecond level, and the internal control logic of the fast flow control subsystem is written in compliance with international standard hard coding.
  5. 5. The industrial control parameter optimization system based on the neural-symbol dual-flow architecture according to claim 1, wherein the fast flow control subsystem performs deterministic safety verification on the proposed value of the process parameter, including a static boundary interval verification for determining whether the proposed value of the process parameter is within a preset safe physical interval and a dynamic change rate verification for determining whether a parameter change rate corresponding to the proposed value of the process parameter is within a preset safe change rate range.
  6. 6. The industrial control parameter optimization system based on the neural symbol double-flow architecture according to claim 5, wherein when the fast flow control subsystem performs static boundary interval verification and dynamic change rate verification, any one of the verification is failed, that is, it is determined that the currently pulled process parameter recommended value is invalid, the safety control target value of the previous scanning period is maintained to run, meanwhile, invalid process parameter recommended values in the intermediate data interaction register module are cleared, and abnormal alarms of corresponding types are triggered, and when both the verification passes, it is determined that the process parameter recommended value is valid, smooth approximation processing is performed on the valid process parameter recommended value, and a final control target value is generated.
  7. 7. The industrial control parameter optimization system based on the neural symbol double-flow architecture according to claim 1, wherein the slow-flow cognitive subsystem cannot directly access an input/output mapping area of the fast-flow control subsystem and cannot directly issue any driving instruction to the physical execution mechanism, and the fast-flow control subsystem only executes a driving signal generated based on a final control target value in a self-scanning period and does not execute any driving instruction issued directly by an external system.
  8. 8. The industrial control parameter optimization system based on the neural symbol double-flow architecture according to claim 1, wherein when the system is deployed in an inventory industrial control system, the non-invasive deployment of the system can be completed by only allocating a corresponding storage address for an intermediate data interaction register module and opening a unidirectional write permission of the corresponding storage address for a slow-flow cognitive subsystem without modifying a core closed-loop adjustment control program code built in a fast-flow control subsystem.
  9. 9. The industrial control parameter optimization system based on the neural symbol double-flow architecture according to claim 1, wherein the operation period of the slow-flow cognitive subsystem is not constrained by the fixed scanning period of the fast-flow control subsystem, and the reasoning delay and the calculation load fluctuation of the slow-flow cognitive subsystem are buffered through the intermediate data interaction registering module, so that the interference to the fixed scanning period of the fast-flow control subsystem is avoided.
  10. 10. The industrial control parameter optimization method based on the neural symbol double-flow architecture is characterized in that the industrial control parameter optimization method based on the neural symbol double-flow architecture is realized based on the industrial control parameter optimization system based on the neural symbol double-flow architecture, and the industrial control parameter optimization method based on the neural symbol double-flow architecture comprises the following steps: Step1, a slow flow cognitive subsystem acquires multi-mode working condition data of an industrial field in a bypass monitoring mode, and performs global optimizing calculation of process parameters by combining built-in historical work order data and process expert experience data to generate a process parameter suggestion value; step 2, the slow flow cognitive subsystem writes the generated technological parameter proposal value into the intermediate data interaction registering module silently, and the writing action does not trigger the hardware interrupt of the fast flow control subsystem; Step 3, actively pulling the currently stored technological parameter proposal value from the intermediate data interaction registering module by the fast flow control subsystem in each fixed scanning period; And 4, finishing the deterministic safety check of the pulled technological parameter recommended value in the same scanning period by the quick control subsystem, processing the technological parameter recommended value into a final control target value after the verification is passed, substituting the final control target value into a built-in closed-loop control algorithm to generate a driving signal, and issuing the driving signal to a corresponding physical executing mechanism to finish the control action, and maintaining the safe control target value of the previous scanning period to operate and triggering an abnormal alarm when the verification is not passed.

Description

Industrial control parameter optimization system and method based on neural symbol double-flow architecture Technical Field The invention relates to the technical field of industrial intelligent control and automation, in particular to an industrial control parameter optimization system and method based on a neural symbol double-flow architecture. Background The industrial development just needs that in the process of converting the manufacturing industry into the intelligent and flexible conversion, the demands of industrial sites on production efficiency improvement, energy consumption refined management, equipment predictive maintenance and process parameter self-adaptive global optimization are obviously improved. The traditional PLC/DCS and other industrial control systems are based on static rule matrixes, conventional linear PID algorithms and fixed time sequence execution flow designs, so that a large number of nonlinear variables caused by raw material fluctuation, environmental parameter change, equipment abrasion and the like in a dynamic manufacturing environment are difficult to cope with, complex multivariable coupling decisions and global real-time optimization cannot be completed, and unstructured implicit knowledge such as experience of operators, maintenance manuals and the like cannot be directly analyzed and utilized, so that industrial data assets are wasted. The technology evolution trend is that a generated artificial intelligence technology represented by an edge side Small Language Model (SLM) has cross-mode data fusion and logic reasoning capabilities, and the cognition capability of the technology is integrated into an auxiliary control decision of an industrial control loop, so that the technology has become a core technology evolution direction in the field of industrial automation. The prior art has the core bottleneck that two technical pain points which cannot be avoided exist in engineering landing in the prior art by introducing a large language model/a small language model into an industrial control loop. Firstly, real-time mismatch and system execution jitter are caused, the hard real-time control requirement of a bottom industrial execution mechanism and the soft real-time characteristic of AI model reasoning have huge time scale difference, the time delay fluctuation of AI reasoning directly penetrates to a physical layer to damage the stability of the original closed-loop control, secondly, the security uncontrollable and override control risk is generated, the inherent 'illusion' problem exists in the generation of AI, the AI directly takes over the architecture of a physical executor in the existing scheme, abnormal out-of-range instructions cannot be intercepted, and equipment damage and safety production accidents are extremely easy to be caused. Disclosure of Invention The invention provides an industrial control parameter optimization system and method based on a neural symbol double-flow architecture, which are used for realizing time scale decoupling of AI cognitive calculation and bottom layer mechanical control, solving the problems of system resource competition and time sequence breakdown caused by the operation of the two same time axes, cutting off a direct driving link of a probabilistic AI model to a bottom layer physical actuator, eliminating override control risks, remolding a bottom layer safety defense line of industrial control, constructing a low-cost, low-risk and non-invasive intelligent upgrading architecture adapting to an old PLC/DCS system, and avoiding high implementation risks of modifying bottom layer core control logic. In order to achieve the above purpose, the invention adopts the following technical scheme: The industrial control parameter optimization system based on the neural symbol double-flow architecture comprises a slow-flow cognitive subsystem, a fast-flow control subsystem and an intermediate data interaction registering module; The slow-flow cognitive subsystem is deployed in edge computing equipment of an industrial field, is internally provided with a small language model subjected to fine adjustment in the industrial vertical field, is used for acquiring multi-mode working condition data of the industrial field in a bypass monitoring mode, and is combined with built-in historical work order data and process expert experience data to execute global optimizing calculation of process parameters with minimum comprehensive production cost as a target so as to generate a process parameter suggestion value; The intermediate data interaction registering module is arranged in an accessible storage area of the fast flow control subsystem and used as a unidirectional communication isolation interface between the slow flow cognition subsystem and the fast flow control subsystem, and is used for receiving and storing a technological parameter proposal value written in the slow flow cognition subsystem, and the writing action of the sl