Neural Network Test Sequences for Hard-to-Reach Computing Modules

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for testing computing devices are inefficient and may fail to detect functional deviations, security vulnerabilities, and unexpected behavior due to inadequate test input coverage, particularly in hard-to-reach components, leading to potential undetected faults and security risks.

Innovation Solution

A neural network-based generator unit, trained on a pre-defined architecture, generates test input sequences to comprehensively test computing modules, utilizing feedback for re-training and optimization to ensure effective coverage and detection of functional and security issues.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional test input selection methods are used, then the testing process is simple to implement, but the test input coverage is insufficient and fails to detect functional deviations in hard-to-reach components

Engineering Contradiction:
Improvedetection capabilityVSAvoidtesting system complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary system comprising a neural network generator and feedback mechanism that bridges the gap between simple test input generation and comprehensive coverage. The generator creates optimized test input sequences while the feedback loop analyzes component activation states, enabling detection of hard-to-reach components without manual intervention in the testing process.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system performs preliminary analysis of the computing device architecture and pre-identifies hard-to-reach components before actual testing begins. The neural network is pre-trained on device specifications and design documents to understand component relationships, enabling it to generate test inputs that specifically target difficult-to-access functional areas from the outset.

Inventive Principle:
Principle #10Preliminary action

2Reliability

If random test input selection is used, then the testing process is fast and simple, but the coverage of test inputs is poor and cannot optimize detection of functional bugs and security vulnerabilities

Engineering Contradiction:
Improvefunctional deviation detectionVSAvoidtest input generation time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent implements a feedback mechanism where the system monitors which components are activated by test inputs and identifies when hard-to-reach components remain inactive. This feedback information is fed back to the neural network generator, which adjusts subsequent test input sequences to specifically target the uncovered components, iteratively improving coverage without exhaustive testing.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The testing system dynamically adapts the test input generation process based on real-time feedback about component activation. The neural network modifies the characteristics of test inputs during execution, changing parameters such as input values, sequences, and patterns to optimize coverage of previously unreachable components while maintaining efficiency.

Inventive Principle:
Principle #15Dynamics

3Reliability

If comprehensive test input coverage is pursued to detect all functional deviations, then detection capability improves, but the complexity and time consumption of the testing process increases significantly

Engineering Contradiction:
Improvecoverage of test inputsVSAvoidtesting efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The system applies partial action by focusing test inputs specifically on identified hard-to-reach components rather than uniformly testing all components. The feedback mechanism determines which components require additional testing attention and directs resources accordingly, achieving comprehensive coverage of critical areas without the overhead of exhaustive testing of all device functions.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP4610883A1Network for generating test input sequences to test the function of computing modules
Publication Date: 2025.09.03 TECH UNIV DARMSTADT
  • EP4610883A1 patent drawingFigure 1
  • EP4610883A1 patent drawingFigure 2
  • EP4610883A1 patent drawingFigure 3

AI summary

To test the operation of a computing module under test, a computer uses (400) a neural network that to generate (410, 450) a test input sequence to test (401) a pre-defined function of the module. The neural network has an internal structure of a generative language model, and it has been trained with a training input sequence of statements that are configured according to the pre-defined architecture of the module. Feedback from testing (401) can serve to optimize the network, such a during re-training (750).