Smart Fuse State Classification via Neural Networks

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for controlling loads and consumers in vehicles, such as relays and melting fuses, are slow and error-prone, failing to adapt effectively to environmental changes and complex current profiles, leading to inadequate protection against overcurrents and short circuits.

Innovation Solution

A computer-implemented method using artificial neural networks, specifically convolutional neural networks and long short-term memory, to classify load current measurements and determine the state of electric and electronic components, enabling smart fuse technology that adapts to environmental changes and provides real-time overload detection without significant processor load.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If machine learning methods are used to determine the state of electric and electronic components, then the adaptability to environmental changes and detection accuracy are improved, but the processor performance requirements and computational load increase

Engineering Contradiction:
Improveadaptability to environmental changesVSAvoidprocessor performance requirements
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The machine learning model is trained offline beforehand with extensive current measurement data covering various environmental conditions and load types. This preliminary training phase separates the computationally intensive learning process from the real-time operation, allowing the model to be deployed as a pre-trained system that requires minimal processing power during actual state determination.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent uses measured current profiles as training data to create a model that copies the characteristic patterns of different states. Instead of performing complex real-time analysis, the system creates a reference model during training that can be quickly compared against new measurements, reducing the computational burden during operation.

Inventive Principle:
Principle #26Copying

2Device complexity

If traditional relays and melting fuses are used for controlling loads, then the device complexity is reduced, but the reaction speed and reliability in detecting overcurrents and short circuits deteriorate

Engineering Contradiction:
Improvedevice complexityVSAvoidreliability in detecting overcurrents and short circuits
Core Design Contradiction:
Device complexityVSReliability

Solution Approach 1:

The patent replaces traditional mechanical relays and melting fuses with an electronic system using current measurements and machine learning-based state determination. This substitution eliminates the slow mechanical operation and error-prone characteristics of traditional devices while maintaining simplicity through software-based control logic.

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

Solution Approach 2:

The system continuously monitors current measurements and automatically determines the state of electric and electronic components without requiring external intervention. The machine learning model self-adjusts and adapts to different conditions, providing autonomous protection against overcurrents and short circuits.

Inventive Principle:
Principle #25Self-service

3Ease of operation

If static thresholds are used for current monitoring, then the ease of operation is improved, but the adaptability to environmental changes and accuracy in complex current profiles deteriorate

Engineering Contradiction:
Improveease of operationVSAvoidadaptability to environmental changes
Core Design Contradiction:
Ease of operationVSAdaptability or versatility

Solution Approach 1:

The patent transitions from static current thresholds to dynamic state determination using machine learning. The system adapts its monitoring criteria based on environmental conditions, load types, and current profile characteristics learned during training. This dynamic approach maintains ease of operation through automated decision-making while significantly improving adaptability to changing conditions.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system changes the monitoring parameters from fixed threshold values to variable state probabilities determined by the machine learning model. The model outputs state determinations based on multiple current measurement parameters including amplitude, frequency, and temporal patterns, allowing flexible adaptation to different environmental conditions without manual threshold adjustment.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3992643B1Methods and systems for determining a state of an arrangement of electric and/or electronic components
Publication Date: 2024.09.25 APTIV TECHNOLOGIES AG
  • EP3992643B1 patent drawingFigure 1
  • EP3992643B1 patent drawingFigure 2
  • EP3992643B1 patent drawingFigure 3

AI summary

A computer implemented method for determining a state of an arrangement of electric and/or electronic components comprises the following steps carried out by computer hardware components: determining a plurality of load current measurements of the arrangement; providing the plurality of load current measurements to a machine learning method; and determining the state of the arrangement based on the plurality of load current measurements using the machine learning method.