Vehicle Sound Evaluation Using Operation Data Deviation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle evaluation systems face challenges in accurately assessing vehicle states due to the rarity of anomaly sound data and the need to differentiate between normal and abnormal vehicle sounds, especially when operating conditions vary.
Innovation Solution
A vehicle evaluation system using a pretrained model trained by supervised learning to generate and evaluate operation data from recorded sound data, incorporating operation patterns and environmental factors, to determine deviations and rank vehicle conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If sound data is used to evaluate vehicle state, then evaluation capability is provided, but accuracy deteriorates due to variation in operating conditions
Solution Approach 1:
The patent introduces operation data as an intermediary variable that mediates between sound data and vehicle state evaluation. The operation data captures operating conditions (accelerator pedal position, brake pedal position, gear position) and serves as a reference to adjust and normalize sound data comparisons, thereby eliminating the negative impact of operating condition variations on evaluation accuracy
Solution Approach 2:
The patent changes the evaluation parameters by incorporating operation data parameters (accelerator pedal position, brake pedal position, gear position) into the evaluation process. By comparing sound data against operation data and adjusting for these parameter variations, the system maintains evaluation accuracy across different operating conditions
2Extent of automation
If anomaly detection is performed using pretrained machine learning models, then detection capability is provided, but reliability deteriorates due to rarity of abnormal sound data
Solution Approach 1:
The patent uses copying by creating operation data copies from multiple sources (vehicle sensors, operator inputs) and using these copies as reference data in the evaluation process. The operation data is copied and stored to serve as a reliable reference for comparing against sound data, enabling the system to overcome the scarcity of abnormal sound data through robust reference data
Solution Approach 2:
The patent implements feedback by using operation data as feedback reference to continuously adjust and refine the evaluation process. The system compares sound data against operation data feedback loops, allowing the pretrained model to maintain reliability by continuously referencing operational context even when abnormal sound data is scarce
Data Source
AI summary
A vehicle evaluation system includes processing circuitry and a memory device. The memory device stores data of a pretrained model. The pretrained model is a model trained by supervised learning to generate operation data from training sound data. The operation data is data indicating changes in operation information in a measurement operation pattern. The processing circuitry is configured to perform a generation process and an evaluation process. The generation process is a process of outputting generated data which is data of operation information generated by inputting the evaluation data to the pretrained model. The evaluation data includes evaluation sound data recorded while operating a target vehicle in the measurement operation pattern. The evaluation process is a process of comparing the operation data with the generated data and evaluating the target vehicle in accordance with the magnitude of a difference therebetween.