Remaining Service Life Prediction from Encoded Sensor Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Technical systems and their components undergo wear processes, leading to inefficiencies and potential failures, necessitating an accurate method to determine remaining service life for timely maintenance and prevention of failures.
Innovation Solution
A computer-implemented method using machine learning systems, specifically autoencoders, variational autoencoders, or normalizing flows, to determine the remaining service life of technical system components by analyzing input signals from sensors, processing them to extract relevant information, and comparing them to historical data to predict the lifespan accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional wear monitoring methods are used, then the system can detect component degradation, but the determination of remaining service life lacks accuracy
Solution Approach 1:
The patent introduces an encoder as an intermediary component that transforms raw sensor input signals into compressed representations. This encoder acts as a mediator between the physical component state and the remaining service life prediction, extracting essential features while filtering out irrelevant information. The encoder representation serves as the key intermediary that enables accurate RUL determination by focusing on the most relevant degradation patterns.
Solution Approach 2:
The patent creates a compressed copy or representation of the component's operational state through the encoder. Instead of directly analyzing raw sensor data, the system generates an encoded representation that captures the essential characteristics of component degradation. This copied representation is then used for RUL prediction, allowing the system to work with a simplified yet informative model of the component state.
2Measurement precision
If comprehensive sensor data is collected to improve prediction accuracy, then the remaining service life can be determined more precisely, but the data processing complexity increases
Solution Approach 1:
The patent extracts only the essential features from comprehensive sensor data through the encoder. Instead of processing all raw sensor information, the encoder selectively extracts the most relevant features that indicate component degradation and remaining service life. This extraction process reduces data complexity while preserving the critical information needed for accurate RUL prediction.
Solution Approach 2:
The patent transforms the parameter representation of sensor data by encoding it into a different parameter space. The encoder changes the parameters from raw sensor readings to compressed representations that are more suitable for RUL prediction. This parameter transformation simplifies the data structure while maintaining the predictive information, reducing processing complexity.
Data Source
AI summary
A computer-implemented method for determining a remaining service life of at least one component of a technical system is disclosed. The method includes (i) determining a first input signal by way of at least one sensor, wherein the first input signal characterizes an operating state of at least the component of the technical system, (ii) determining a first representation of the first input signal by way of an encoder of a first machine learning system, and (iii) determining the remaining service life on the basis of the first representation and on the basis of a provided plurality of second representations, wherein the plurality of second representations is determined on the basis of a plurality of second input signals by way of the encoder and a corresponding remaining service life is assigned to each second representation.