Filter Debugging Using Reinforcement Learning Policy Network

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The high labor and time costs associated with manually debugging filters, such as ceramic dielectric filters, result in low efficiency, as each filter requires repeated adjustments by skilled workers to meet performance standards.

Innovation Solution

A reinforcement learning-based policy network is trained to determine the debugging scheme for filters based on hole parameters and index values, enabling intelligent debugging by automating the process and replacing manual adjustments.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If manual debugging by skilled workers is used, then filter performance requirements are met, but labor cost and time cost are high

Engineering Contradiction:
Improvefilter performanceVSAvoiddebugging efficiency
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The patent replaces the mechanical manual debugging system with an automated system comprising a mechanical arm, simulation system, and policy network. The mechanical arm automatically performs polishing operations on filter holes based on debugging schemes generated by the policy network, eliminating the need for manual intervention while maintaining filtering performance requirements.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent implements a self-service debugging system where the filter itself provides information about its current state through the simulation system, which then automatically generates debugging schemes. The system autonomously determines what adjustments are needed without external human input, achieving self-debugging capability that improves efficiency while meeting performance standards.

Inventive Principle:
Principle #25Self-service

2Reliability

If repeated manual debugging is performed, then filter index requirements are satisfied, but time consumption increases

Engineering Contradiction:
Improvefilter index complianceVSAvoiddebugging time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent employs a pre-trained policy network that has learned optimal debugging strategies through reinforcement learning. This pre-trained model can immediately generate effective debugging schemes without requiring iterative trial-and-error, significantly reducing the time needed to achieve compliance with filter index requirements while ensuring reliable performance.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements a closed-loop feedback system where the simulation system continuously monitors filter performance indices and provides real-time feedback to the policy network. This feedback mechanism enables the system to automatically adjust debugging schemes based on current filter state, ensuring index compliance is achieved efficiently through iterative but automated adjustments rather than repeated manual debugging.

Inventive Principle:
Principle #23Feedback

3Productivity

If automated debugging is implemented, then debugging efficiency improves, but system complexity increases

Engineering Contradiction:
Improvedebugging efficiencyVSAvoidsystem complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent employs a universal policy network that can handle multiple filter types and debugging scenarios through reinforcement learning training on diverse datasets. This single multi-functional system replaces what would otherwise require multiple specialized systems, achieving high debugging efficiency across different filter configurations without proportionally increasing system complexity.

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

Solution Approach 2:

The patent introduces a simulation system as an intermediary between the physical filter and the policy network. This virtual simulation environment serves as a mediator that translates complex physical filter states into simplified representations that the policy network can process efficiently, reducing the computational complexity while maintaining debugging effectiveness.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP3832792B1Filter debugging method, device, electronic apparatus, readable storage medium and computer program product
Publication Date: 2023.04.26 BEIJING BAIDU NETCOM SCI & TECH CO LTD
  • EP3832792B1 patent drawingFigure 1
  • EP3832792B1 patent drawingFigure 2~3
  • EP3832792B1 patent drawingFigure 4~5

AI summary

A filter debugging method, a device, an electronic apparatus, a readable storage medium and a computer program product are provided. The filter debugging method includes: step S1: inputting a current hole parameter and a current index value of a filter into a policy network which is pre-trained; step S2: determining, by the policy network, a target hole to be polished of the filter, according to the current hole parameter and the current index value of the filter; step S3: controlling a mechanical arm to polish the target hole of the filter; and step S4: determining whether the filter is qualified according to an index value of the polished filter; in a case that the filter is qualified, ending a process including the steps S1 to S4; in a case that the filter is unqualified, performing the steps S1 to S4 circularly until the filter is qualified.