Neural Network Development System for Vehicle Adaptability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for developing devices, such as vehicle components, rely heavily on artificial neural networks that are dependent on training data and weighting, making it challenging to combine different interests and customize solutions for various vehicle models and configurations in series production.
Innovation Solution
A method utilizing a development system with multiple artificial neural networks trained on different data sets, an evaluation network for result assessment, and a standardization level to identify overlaps, allowing for the creation of competing development results and selecting the best implementation, while enabling continuous improvement through feedback and pre-development processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a single artificial neural network is used for device development, then the development process is simple, but the results are highly dependent on training data and random weightings, limiting adaptability
Solution Approach 1:
The development system is segmented into multiple independent artificial neural networks (first network, second network, evaluation network), each trained on different datasets. This segmentation allows each network to specialize in different aspects of device development while maintaining overall system adaptability across various vehicle models and configurations.
Solution Approach 2:
The multi-network system is designed to serve multiple functions: the first and second networks generate different development outcomes, while the evaluation network assesses these outcomes against predefined criteria. This universal system handles diverse vehicle model requirements through a unified multi-network architecture.
2Adaptability or versatility
If multiple artificial neural networks are used to solve the same task, then adaptability improves, but the device complexity and computational requirements increase
Solution Approach 1:
The evaluation network acts as an intermediary between the first and second networks and the final decision-making process. It receives development outcomes from both networks, evaluates them against predefined criteria, and outputs documentation data that guides the selection of the best development result, thereby managing the complexity of coordinating multiple networks.
Solution Approach 2:
The system implements feedback through the evaluation network, which assesses development outcomes and provides documentation data that can be used to refine and improve subsequent development processes. This feedback mechanism helps manage complexity by providing structured information about the performance of multiple networks.
3Manufacturing precision
If development results are evaluated based on predefined criteria, then manufacturing precision improves, but the loss of time for evaluation increases
Solution Approach 1:
The evaluation network is pre-configured with predefined evaluation criteria before the development process begins. This preliminary setup allows the network to quickly assess development outcomes against established standards without requiring time-consuming ad-hoc evaluation, thereby reducing evaluation time while maintaining precision.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method (100) for supporting the development of a device (1) by means of a development process (200) in a development system (2) comprising at least one development level (10) comprising at least one artificial neural first network (11) trained by a first data set (211) of training data, and an artificial neural second network (12) trained by a second data set (212) of training data, and an evaluation level (20) comprising at least one neural evaluation network (21) for evaluating development results of the development level (10). The invention further relates to a computer program product and a development system (2).