Plant Failure Prediction Model Selection for Aging Equipment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing failure prediction models for large plants, such as power generation or chemical plants, face challenges in accurately predicting failures due to complexity, lack of quality data, and incomplete failure detection/diagnosis information, leading to decreased accuracy over time.
Innovation Solution
A failure prediction model generating apparatus and method that selects an optimal failure prediction model from a plurality of models using data collected from the plant, allowing for the generation of a new model based on the prediction result, incorporating both physics-based and data-based models to improve prediction accuracy and reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a failure prediction model is generated initially based on domain knowledge, then the prediction can be made, but the accuracy decreases over time due to plant degradation and part replacement
Solution Approach 1:
The system dynamically adapts the failure prediction model by selecting between physics-based models and data-based models based on current plant conditions. The model selection is not static but changes according to the accumulated operating time and degradation state, allowing the system to maintain high prediction accuracy throughout the plant's operational lifecycle
Solution Approach 2:
The system changes the parameters of the prediction approach by switching between different modeling paradigms (physics-based vs. data-based) depending on the operational phase. Early in operation, physics-based models with domain knowledge are used, while later when degradation patterns are established, data-based models trained on actual plant data provide better predictions
2Reliability
If domain knowledge is used for failure prediction, then the model can be constructed, but accurate prediction is hindered by plant complexity, lack of quality data, and incomplete failure detection information
Solution Approach 1:
The system segments the prediction task into two distinct approaches: physics-based modeling for understanding fundamental failure mechanisms and data-based modeling for capturing complex empirical patterns. This segmentation allows each approach to address specific aspects of plant complexity without being overwhelmed by the other
Solution Approach 2:
The system introduces an intermediary model selection mechanism that bridges the gap between physics-based models and data-based models. This intermediary selects the appropriate modeling approach based on current conditions, effectively mediating between the limitations of domain knowledge and the requirements for accurate prediction in complex plant environments
Data Source
AI summary
A failure prediction model generating apparatus and method thereof are provided. The failure prediction model generating apparatus includes a memory configured to store a plurality of failure prediction models derived previously; and a processor configured to predict a failure of the plant, wherein the processor is configured to collect data measured from the plant, select at least one failure prediction model from among the plurality of failure prediction models using the collected data, and predict a failure of the plant using the selected failure prediction model.