Neural-Network PID Tuning for Stable Fluid Quantity Control

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current control systems for fluid physical variables, such as pressure and temperature, face challenges in achieving stable and precise control, particularly in industrial and medical applications, due to complexity in commissioning and stability issues, which limits their widespread adoption.

Innovation Solution

A control device comprising a control unit and an optimization unit, where the optimization unit adapts control parameters based on actual values of physical variables, using an artificial neural network to enhance the control parameters' precision and stability without the need for extensive training, allowing for adaptive and efficient control.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If adaptive control with neural networks is implemented to improve control precision and stability, then manufacturing precision and reliability are improved, but device complexity and commissioning difficulty increase

Engineering Contradiction:
Improvecontrol precisionVSAvoidsystem complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The control system is segmented into distinct functional modules: a control unit that executes control algorithms and an optimization unit that adapts control parameters. This segmentation allows the complex adaptive control functionality to be distributed across separate components, making the system more manageable and easier to commission while maintaining high control precision through the optimization unit's neural network-based parameter adaptation.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The optimization unit performs preliminary adaptation of control parameters using neural networks before the control unit executes the actual control operations. By pre-optimizing parameters based on process characteristics, the system achieves high manufacturing precision without requiring complex real-time adjustments during operation, thereby reducing overall system complexity.

Inventive Principle:
Principle #10Preliminary action

2Reliability

If adaptive control with neural networks is implemented to improve control precision and stability, then reliability is improved, but ease of operation deteriorates due to complex commissioning

Engineering Contradiction:
Improvecontrol stabilityVSAvoidcommissioning ease
Core Design Contradiction:
ReliabilityVSEase of operation

Solution Approach 1:

The optimization unit implements self-service adaptive control by automatically adjusting control parameters using neural networks based on process feedback. This self-adjustment capability ensures reliable and stable control without requiring manual commissioning or continuous operator intervention, thereby maintaining high reliability while improving ease of operation through automated parameter optimization.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system incorporates feedback mechanisms where the optimization unit continuously monitors control performance and adjusts parameters accordingly. This closed-loop feedback ensures reliable control stability while automating the commissioning process, as the system learns and adapts parameters automatically based on observed process behavior rather than requiring manual tuning.

Inventive Principle:
Principle #23Feedback

3Device complexity

If traditional control methods are used to maintain simple system structure, then device complexity is reduced, but manufacturing precision and control accuracy deteriorate

Engineering Contradiction:
Improvesystem simplicityVSAvoidcontrol accuracy
Core Design Contradiction:
Device complexityVSManufacturing precision

Solution Approach 1:

The control unit is designed with multi-functionality, capable of executing both traditional control algorithms and adaptive optimization routines. This universal design allows the system to maintain simple operational structure while incorporating advanced precision control capabilities through the optimization unit, achieving high manufacturing precision without significantly increasing apparent system complexity.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The optimization unit acts as an intermediary between the simple control unit and the complex adaptive control requirements. It translates process requirements into optimized control parameters that the simple control unit can execute, thereby achieving high manufacturing precision through the intermediary's parameter optimization while maintaining the simplicity of the main control structure.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP3825788B1Control device, control system and control method for regulating a physical quantity of a fluid
Publication Date: 2022.11.09 ASCO NUMATICS
  • EP3825788B1 patent drawingFigure 1
  • EP3825788B1 patent drawingFigure 2A~2B
  • EP3825788B1 patent drawingFigure 3~5

AI summary

The present disclosure relates to a control device (150) for controlling a physical quantity of a fluid, comprising a control unit (10), e.g., a PID controller, which receives at least one first actual value (14) and at least one first setpoint (12) of a physical quantity of the fluid, compares the first actual value (14) and the first setpoint (12), and, based on this comparison, calculates a manipulated variable (16); and comprising an optimization unit (70), which includes an artificial neural network and which receives signals describing at least one physical quantity of the fluid in the form of at least one actual value as the first actual value (14), adjusts a control parameter (74), e.g., the PID controller parameters, based on the at least one actual value, and passes it to the control unit (10), wherein the control unit (10) calculates the manipulated variable (16) based on the optimized control parameter (74) in order to output an optimized manipulated variable (16).Thus, a neural network itself is not used as the controller; instead, the control parameters (74) of the control unit (10) are adaptively adjusted by the optimization unit (70). This eliminates the need for the time-consuming training of a controller with a neural network.