Soundness Diagnosis Using Operator-Normalized Feature Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional soundness diagnosis techniques for manually operated devices, such as railway vehicle brakes, suffer from decreased accuracy due to individual variations in sensor data from different operators.
Innovation Solution
A soundness diagnosis apparatus that acquires operating data, generates feature amount data based on physical characteristics, performs model learning of a normal state using machine learning, and visualizes diagnosis results to reduce the impact of individual variations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If sensor data from multiple operators is used for device diagnosis, then more data is available for analysis, but individual variations reduce diagnosis accuracy
Solution Approach 1:
The patent extracts and removes individual variation components from the sensor data through statistical processing. By separating the operator-specific variations from the actual device state information, the system retains useful diagnostic data while eliminating the harmful individual variations that reduce accuracy.
Solution Approach 2:
The patent transforms the raw sensor data by changing its statistical parameters - specifically by calculating deviation values from operator-specific baseline statistics. This parameter transformation converts data with individual variations into standardized deviation data that eliminates those variations while preserving diagnostic information.
2Ease of manufacture
If conventional diagnosis methods are used for manually operated devices, then implementation is simple, but accuracy decreases due to individual operator variations
Solution Approach 1:
The patent performs preliminary statistical processing to establish operator-specific baseline data before actual diagnosis occurs. By pre-calculating mean values and standard deviations for each operator and storing them as reference data, the system prepares the necessary correction mechanisms in advance, making the subsequent diagnosis process both accurate and computationally efficient.
Solution Approach 2:
The patent introduces statistical parameters (mean value and standard deviation) as intermediary elements between the raw sensor data and the diagnosis result. These intermediaries serve as mediators that transform operator-specific data into standardized deviation values, enabling accurate comparison across different operators while maintaining implementation feasibility.
Data Source
AI summary
A soundness diagnosis apparatus diagnoses soundness of a device includes: a data loading unit that acquires operating data of the device for a diagnosis target period; a feature amount data generation unit that samples, as sample data, a target data segment for feature amount data from the operating data on the basis of a physical characteristic of the device, and generates the feature amount data by using the sample data; an inference unit that performs soundness diagnosis on the feature amount data that is sequentially generated by using a learned model obtained through model learning of a normal state of the device; and a visualization unit that visualizes a transition of a soundness diagnosis result obtained by the inference unit.