Industrial AI Drift Mitigation Through Context-Based Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Industrial plants experience data drift due to variations in operating conditions, leading to AI model degradation and resulting in poor product quality, equipment failures, and plant shutdowns, with conventional data sampling methods often failing to adequately address this issue.
Innovation Solution
A system and method that utilizes a processor to monitor process parameters, detect drift, determine context and context-based sampling strategies, and train a second AI model to mitigate data drift, enhancing the AI model's ability to control industrial processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data sampling is performed to collect and learn about new characteristics of process parameters, then AI model performance can be maintained, but inaccurate sampling may lead to inadequate and/or negative training
Solution Approach 1:
The patent introduces drift context and process context as intermediary elements that mediate between the raw process parameters and the AI model training process. These contexts act as mediators that guide the sampling strategy selection, ensuring that the sampling is both accurate and appropriate for the specific drift scenario being encountered.
Solution Approach 2:
The patent changes the parameters of the sampling process by selecting from multiple sampling strategies based on drift context and process context. Instead of using a fixed sampling approach, the system dynamically adjusts sampling parameters (which strategy to use, what to sample, how much to sample) based on the detected drift characteristics and current process state.
2Reliability
If AI models are trained on historical process parameters, then initial model performance is achieved, but model performance degrades when operating conditions change
Solution Approach 1:
The patent performs preliminary action by detecting drift and determining appropriate sampling strategies before the AI model performance significantly degrades. The system proactively identifies when drift is occurring and prepares the appropriate training data through context-based sampling, preventing rather than just reacting to model degradation.
Solution Approach 2:
The patent implements a feedback loop where the system continuously monitors process parameters, detects drift when it occurs, determines drift context and process context, selects appropriate sampling strategies, and retrains the model accordingly. This closed-loop feedback mechanism enables the model to adapt to changing operating conditions while maintaining performance.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
The present disclosure discloses a system and a method for mitigating data drift in an industrial plant. The method includes monitoring, by a processor (202), one or more process parameters associated with an industrial plant (102). The method includes detecting, by processor (202), a drift in one or more process parameters based on a deviation from one or more predefined process parameters. The method includes determining, by processor (202), one or more drift context and process context based on drift and one or more process parameters. The method includes determining, by processor (202), sampling strategy from plurality of sampling strategies based on one or more drift and process context for sampling one or more process parameters using first Artificial Intelligence (AI) model (216). The method includes training, by processor (202), second Al model (302) based on sampling strategy for mitigating data drift.