Sensor Maintenance Data Labeling for Deterioration Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The high cost of maintenance for sensors in industrial plants due to the need for specialized workers and periodic maintenance, as well as the increased cost of manually preparing teacher data for machine learning-based deterioration determination, is a significant challenge.
Innovation Solution
An information processing device and method that acquires measured and characteristic data from sensors, generates teacher data by associating settling time with historical data, and uses machine learning to determine necessary maintenance, reducing the need for specialized workers and manual data preparation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If periodic maintenance is performed on all sensors in a predetermined cycle, then sensor reliability is maintained, but maintenance costs increase due to unnecessary maintenance on sensors that do not require it
Solution Approach 1:
The patent changes the maintenance parameter from fixed periodic intervals to variable intervals based on actual sensor deterioration state. The determination device calculates deterioration degrees using machine learning models and adjusts maintenance timing accordingly, allowing sensors to be maintained only when necessary rather than following a rigid schedule for all sensors
Solution Approach 2:
The system enables sensors to essentially self-diagnose their own state through the determination device that calculates deterioration degrees and predicts remaining useful life. This automated assessment replaces manual inspection and allows the system to identify which sensors need maintenance without human intervention, reducing unnecessary maintenance actions
2Measurement precision
If machine learning is used to determine sensor deterioration, then maintenance precision is improved, but the complexity of data preparation increases due to manual teacher data creation
Solution Approach 1:
The patent performs preliminary actions by automatically collecting and preparing training data before machine learning model execution. The determination device gathers historical sensor data and automatically associates it with maintenance outcomes to create teacher data, eliminating the need for manual data preparation and enabling seamless deployment of machine learning models
Solution Approach 2:
The determination device acts as an intermediary between raw sensor data and machine learning models. It automatically processes, cleans, and structures data into the required teacher data format, serving as a bridge that eliminates the manual data preparation bottleneck and enables straightforward integration of machine learning capabilities
3Measurement precision
If specialized maintenance workers are deployed to measure characteristic data, then measurement precision is improved, but labor costs increase
Solution Approach 1:
The patent replaces the mechanical system of specialized workers physically measuring characteristic data with an automated determination device that calculates deterioration degrees from operational data. This substitution eliminates the need for specialized labor while maintaining or improving measurement precision through consistent, objective algorithmic assessment
Solution Approach 2:
The system enables self-assessment of sensor characteristics by automatically calculating deterioration degrees from operational data without requiring external specialized workers. The determination device performs what previously required expert intervention, allowing the system to self-diagnose sensor health status using readily available operational parameters
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An information processing device according to one aspect of the present invention includes a first acquirer configured to acquire measured data of a sensor, a second acquirer configured to acquire characteristic data of the sensor, the characteristic data having been acquired by maintaining the sensor, and a first generator configured to generate teacher data in which the acquired characteristic data is associated as label information with the acquired measured data.