Mitigating Spillover False Alarms in ML Prognostic Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine-learning-based prognostic-surveillance systems often generate spillover false alarms due to unexpected dependencies among input and output signals, leading to unnecessary maintenance and potential improper human actions, which are not related to the quality of the monitored components or sensors.
Innovation Solution
A system that determines susceptibility to spillover false alarms by training an inferential model on time-series signals, adding degradation, and computing moving-window sequential probability ratio test (SPRT) tripping frequencies to detect increased ratios, then removes causal signals and retrains the model to mitigate spillover false alarms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If an inferential model is trained to detect anomalies in time-series signals, then the system can identify incipient anomalies and enable preventive maintenance, but the model generates spillover false alarms that trigger unnecessary maintenance operations
Solution Approach 1:
The patent segments the anomaly detection process by introducing a separate spillover detection mechanism that operates independently from the primary anomaly detection model. This secondary model specifically identifies spillover false alarms by analyzing patterns characteristic of spillover events, allowing the system to distinguish between true anomalies and spillover false alarms, thereby reducing unnecessary maintenance operations while maintaining reliable anomaly detection
Solution Approach 2:
The patent introduces an intermediary spillover detection model that acts as a mediator between the primary anomaly detection model and the maintenance decision-making process. This intermediary model analyzes the outputs of the primary model and filters out spillover false alarms before they trigger unnecessary maintenance actions, effectively resolving the contradiction between maintaining high detection accuracy and avoiding false alarm-induced unnecessary maintenance
2Measurement precision
If higher-accuracy sensors are deployed to improve monitoring precision, then the quality of sensor data increases, but spillover false alarms increase as well
Solution Approach 1:
The patent converts the harmful effect of spillover false alarms into a beneficial diagnostic opportunity by training a specialized spillover detection model to recognize and identify spillover patterns. Instead of simply filtering out high-accuracy sensor data that causes spillover, the system leverages the precise measurements from these sensors to train a model that can distinguish spillover-induced anomalies from genuine failures, thereby maintaining measurement precision while eliminating the harmful spillover effect
3Reliability
If components are replaced in response to spillover false alarms, then the system attempts to address potential failures, but the asset continues to generate spillover false alarms from newly replaced components
Solution Approach 1:
The patent applies preliminary action by implementing spillover detection and mitigation before maintenance decisions are made. The spillover detection model analyzes anomaly detections in real-time and identifies those caused by spillover effects before they trigger maintenance actions. This preliminary filtering prevents unnecessary component replacements, saving time and resources while avoiding the cycle of repeated unnecessary maintenance on newly replaced components
Data Source
AI summary
The disclosed embodiments relate to a system that determines whether an inferential model is susceptible to spillover false alarms. During operation, the system receives a set of time-series signals from sensors in a monitored system. The system then trains the inferential model using the set of time-series signals. Next, the system tests the inferential model for susceptibility to spillover false alarms by performing the following operations for one signal at a time in the set of time-series signals. First, the system adds degradation to the signal to produce a degraded signal. The system then uses the inferential model to perform prognostic-surveillance operations on the time-series signals with the degraded signal. Finally, the system detects spillover false alarms based on results of the prognostic-surveillance operations.