Anomaly Stream Generation for Predictive Maintenance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional system monitoring systems in industrial environments are limited in providing actionable intelligence from anomalies detected in sensor data, as they lack the necessary information to convert anomaly detection into predictive insights for variables of interest such as time to failure, efficiency, or operating conditions.
Innovation Solution
The technology generates anomaly streams from sensor data, which are then used to construct supervised learning models that predict variables of interest by combining statistical and predictive anomalies, improving predictive accuracy through filtering and feature engineering, and extending these methods to multiple devices for collective anomaly measures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If conventional anomaly detection systems are used to detect anomalies in sensor data, then anomalies can be identified, but the systems lack the capability to convert anomaly detection into actionable predictive intelligence
Solution Approach 1:
The patent introduces an intermediary processing layer that transforms raw anomaly data into enriched anomaly objects containing contextual information, statistical measures, and predictive attributes. This intermediary transformation enables the system to bridge the gap between simple anomaly detection and actionable predictive intelligence by adding meaningful information layers without fundamentally changing the detection mechanism.
Solution Approach 2:
The system performs preliminary enrichment of anomaly data during the detection phase, pre-computing statistical measures, contextual relationships, and predictive attributes before the anomalies are fully utilized. This preliminary action ensures that when anomalies are detected, they already contain the informational content needed for predictive analysis, eliminating the need for separate information-gathering steps.
2Speed
If simple anomaly detection methods are applied to sensor data, then detection speed is maintained, but the ability to predict variables of interest such as time to failure and efficiency is insufficient
Solution Approach 1:
The patent segments the anomaly processing pipeline into distinct functional stages: detection, enrichment, and prediction. Each stage operates independently with optimized algorithms appropriate to its specific task. This segmentation allows the detection phase to maintain high speed while the enrichment and prediction phases enhance measurement precision without creating bottlenecks that would slow down the overall system.
Solution Approach 2:
The system dynamically adjusts processing parameters based on the type of anomaly detected and the specific prediction task at hand. By changing parameters such as the level of enrichment, the complexity of predictive models applied, and the thresholds for different anomaly types, the system optimizes the balance between detection speed and predictive accuracy for each specific scenario.
Data Source
AI summary
A method for predicting variables of interest related to a system includes collecting one or more sensor streams over a time period from sensors in the system and generating one or more anomaly streams for the time period based on the sensor streams. Values for variables of interest for the time period are determined based on the sensor streams and the anomaly streams. Next, a time-series predictive algorithm is applied to the (i) the sensor streams, (ii) the anomaly streams, and (iii) the values for the variables of interest to generate a model for predicting new values for the variables of interest. The model may then be used to predict values for the variables of interest at a time within a new time period based on one or more new sensor streams.