Neural Network Configuration via Changeability Index
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In industrial applications, neural networks used in collaborative engineering environments face challenges due to their black box nature, making it difficult to verify and analyze their accuracy and manage their configuration across multiple partners, leading to potential quality and integrity issues in machine learning components.
Innovation Solution
A method is introduced to configure neural networks by determining a changeability index based on system architecture, using a restriction matrix to control modifications of neural network parameters, ensuring that safety-critical features are not compromised, and providing a collaborative configuration environment through an Integrated Development Environment (IDE) that integrates with lifecycle and manufacturing execution software.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If neural networks are used in industrial applications, then productivity and automation are improved, but verification and analysis of accuracy becomes difficult due to black box nature
Solution Approach 1:
The patent introduces an intermediary system that acts as a mediator between the neural network and the verification process. This intermediary captures and represents the neural network's decision-making logic in a verifiable format, enabling analysis without exposing the internal black box structures. The intermediary translates opaque neural network operations into interpretable representations that can be verified against requirements.
2Adaptability or versatility
If neural networks are configured in collaborative engineering environments, then adaptability and versatility are improved, but managing configuration and determining adjustability becomes unclear
Solution Approach 1:
The patent segments the neural network configuration into distinct, manageable components with defined boundaries. Each segment represents a specific functional unit or parameter set that can be independently configured, tracked, and verified. This segmentation enables clear assignment of responsibilities among collaborators and simplifies the tracking of modifications across the distributed engineering team.
Solution Approach 2:
The patent implements dynamic configuration management that adapts to collaborative needs. The system dynamically tracks which parameters are adjustable and which are fixed, updating this information based on the current collaboration state and verification requirements. This dynamic approach allows the configuration to evolve while maintaining clarity about adjustability constraints.
3Manufacturing precision
If neural network parameters are modified to improve accuracy, then manufacturing precision is improved, but reliability of safety-critical features may be compromised
Solution Approach 1:
The patent applies local quality by treating different parameters of the neural network differently based on their impact on safety-critical features. Critical parameters are protected with stricter modification constraints and verification requirements, while non-critical parameters allow greater flexibility. This localized approach to parameter management enables accuracy improvements in non-critical areas without compromising safety-critical functionality.
Data Source
AI summary
A neural network model, method and modelling environment for configuring neural networks is disclosed. The method of configuring at least one neural network (120) used in an industrial environment, wherein the industrial environment (112, 114, 116) comprises at least one industrial system with one or more hardware and software components, the method comprises receiving the neural network, wherein the neural network (120) is trained based on at least one trained dataset associated with a system architecture of the industrial system; determining modified-parameters of the neural network (120) based on the training; defining a changeability index for the trained neural network (172) based on comparison of the modified-parameters with predefined-parameters of the neural network (120); and configuring the neural network (120, 172) based on the changeability index.