Hybrid Predictive Maintenance Modeling for Explainable Failure Timing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional predictive maintenance techniques face challenges in achieving accurate predictions due to uncertainties in material strength, load calculations, and structural dimensions, requiring large amounts of training data and significant engineering effort, and often lack accountability in explaining prediction results.
Innovation Solution
An information processing device generates two models, a prediction model and a physical model, using sub-libraries that incorporate domain and engineering knowledge, allowing for more accurate predictions with less data and improved accountability by combining nonlinear basis functions and adjusting generation probabilities and hyperparameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional predictive maintenance techniques are used, then predictions can be made for system maintenance timing, but prediction accuracy is poor due to uncertainties in material strength, load calculations, and structural dimensions
Solution Approach 1:
The patent combines multiple prediction models (physical models, data-driven models, and hybrid models) into an integrated predictive maintenance system. This merging of different modeling approaches allows the system to leverage the strengths of each model type while compensating for their individual weaknesses, thereby improving both prediction accuracy and reliability despite uncertainties in material properties and load calculations
Solution Approach 2:
The patent employs composite modeling strategies that integrate different types of models (physics-based and data-driven) similar to how composite materials combine different substances to achieve superior properties. This composite approach creates a more robust prediction framework that maintains high accuracy and reliability even when individual model components have uncertainties
2Measurement precision
If conventional predictive maintenance techniques are used, then life prediction can be performed, but large amounts of training data are required
Solution Approach 1:
The patent changes the parameters and structure of prediction models to reduce data requirements. By incorporating physics-based constraints and domain knowledge into the modeling framework, the system can achieve accurate predictions with fewer training data points, as the physical laws and engineering principles provide additional information that compensates for limited data
Solution Approach 2:
The patent introduces physical models and domain knowledge as intermediaries between raw data and predictions. These intermediary models encode expert knowledge and physical principles that guide the prediction process, reducing the amount of training data needed while maintaining or improving prediction accuracy
3Measurement precision
If conventional predictive maintenance techniques are used, then system failure prediction can be made, but significant engineering effort is required
Solution Approach 1:
The patent creates a universal predictive maintenance platform that can handle multiple types of systems and failure modes through a standardized framework. This multi-functional system reduces engineering effort by providing reusable models, templates, and procedures that can be adapted to different applications without requiring extensive custom development for each case
Solution Approach 2:
The patent segments the predictive maintenance system into modular components (data collection, preprocessing, model selection, prediction, and validation modules). This segmentation allows engineers to work on specific modules independently and reuse completed modules across different projects, significantly reducing overall engineering effort while maintaining prediction accuracy
4Measurement precision
If conventional predictive maintenance techniques are used, then maintenance timing can be predicted, but accountability in explaining prediction results is lacking
Solution Approach 1:
The patent implements feedback mechanisms that provide explanations and justifications for prediction results. The system feeds back information about which models contributed to predictions, what factors influenced the results, and how confident the system is in its predictions. This feedback loop maintains accountability by making the prediction process transparent and interpretable while preserving prediction accuracy
Data Source
AI summary
According to one embodiment, an information processing device includes a memory and one or more processors coupled to the memory. The one or more processors are configured to: generate, by machine learning using time-series data of a variable for a phenomenon related to an abnormality in a system to be monitored, a prediction model for predicting an indicator used to identify the a timing of system maintenance and a physical model for predicting the variable; and perform either one of a first prediction process using the physical model that is learned using the indicator predicted by the prediction model and a second prediction process of correcting the indicator predicted by the prediction model by using the variable predicted by the physical model.