Mechanical Equipment State Transition Data for Malfunction Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to generate a malfunction prediction model with high prediction accuracy for mechanical equipment, especially when only a few cases of malfunction data are available, due to the difficulty in selecting and obtaining suitable learning data.
Innovation Solution
A control method and apparatus that extract measurement values corresponding to transitions from a normal state to a malfunction state, calculate a separation degree for feature values, and use these values to select data for machine learning, enabling the generation of a malfunction prediction model even with limited malfunction data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional machine learning methods are used with limited malfunction data, then the model cannot achieve high prediction accuracy, but increasing maintenance frequency to gather more data decreases operation rate
Solution Approach 1:
The patent extracts and focuses only on the critical transition period data when equipment moves from normal to malfunction state, rather than using all available data. This extraction of essential features allows the model to achieve high prediction accuracy with limited malfunction cases, resolving the contradiction between prediction accuracy and operation rate.
Solution Approach 2:
The patent performs preliminary identification of the transition period between normal and malfunction states, then uses this identified period to select appropriate learning data. This preliminary action enables the system to prepare accurate prediction models without requiring excessive maintenance operations, thus maintaining high operation rate while achieving accurate predictions.
2Device complexity
If all measurement data is used for machine learning, then redundant data increases processing complexity, but selecting only relevant data requires sophisticated selection criteria
Solution Approach 1:
The patent segments the measurement data into three distinct periods: normal period, transition period, and malfunction period. By dividing the data in this manner, the system can easily identify and select only the relevant transition period data for machine learning, reducing processing complexity while ensuring high learning data quality without requiring sophisticated selection criteria.
3Reliability
If maintenance operations are performed frequently to prevent malfunctions, then preventive safety increases, but operation rate decreases due to machine stoppage
Solution Approach 1:
The patent implements a feedback mechanism where the machine learning model continuously evaluates equipment state using measurement data, particularly focusing on the transition period before malfunction. This feedback allows the system to predict malfunctions accurately and perform maintenance only when necessary, thereby maintaining high preventive safety while minimizing maintenance frequency and preserving operation rate.
Data Source
AI summary
A control apparatus includes a controller. The controller is configured to obtain a measurement value of a state of mechanical equipment corresponding to a period in which the mechanical equipment reaches a second state from a first state, extract at least one predetermined feature value by using the measurement value, and extract data for machine learning from data of the at least one predetermined feature value on a basis of a separation degree for distinguishing the first state and the second state from each other.