Problem-Alert Aggregation via Asset Representation Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current predictive maintenance and failure detection systems in industrial plants produce excessive and unreliable alerts, leading to high misdetection rates and inefficient problem identification due to complex system interconnectivity and the need for extensive expert knowledge, resulting in low detection rates and increased maintenance costs.
Innovation Solution
A method for problem-alert aggregation that utilizes an asset representation, including interconnected graphs to group and score alerts, reducing alert rates and improving alert reliability by localizing issues within complex physical systems, allowing for more effective investigation and maintenance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If machine-learning algorithms process sensor readings to detect problems, then problem detection capability is improved, but false alarm rate increases
Solution Approach 1:
The patent introduces an intermediary layer between sensor data and alerts that uses asset representation models to contextualize sensor readings. This intermediary layer translates raw sensor data into meaningful asset states, reducing false alarms by filtering out normal variations that don't represent actual problems.
Solution Approach 2:
The system dynamically adjusts detection parameters based on asset representation and historical data. By changing thresholds and detection sensitivity based on contextual information from the asset model, the system maintains high problem detection capability while reducing false alarm rates.
2Reliability
If thresholds are set low to detect more problems, then detection rate is improved, but alert volume increases excessively
Solution Approach 1:
The patent merges multiple related alerts into single consolidated events by grouping them according to asset representation. Instead of treating each sensor alert independently, the system combines alerts that relate to the same underlying asset problem, significantly reducing alert volume while maintaining high detection rates.
Solution Approach 2:
The system segments alerts by asset components and relationships defined in the asset representation model. This segmentation allows the system to process and consolidate alerts efficiently, grouping related alerts together while maintaining the ability to detect individual problems.
3Productivity
If thresholds are set high to reduce alerts, then alert volume is reduced, but crisis detection capability deteriorates
Solution Approach 1:
The system performs preliminary analysis using asset representation models to evaluate the significance of alerts before they reach maintenance teams. By pre-filtering and prioritizing alerts based on their relationship to asset criticality and historical patterns, the system reduces volume while ensuring crises are not missed.
Solution Approach 2:
The system uses feedback from asset representation and historical maintenance data to continuously refine threshold settings. This feedback mechanism ensures that thresholds remain optimized for both reducing alert volume and maintaining crisis detection capability.
4Measurement precision
If maintenance teams investigate all alerts manually, then alert investigation thoroughness is improved, but time consumption increases
Solution Approach 1:
The patent introduces an intermediary system that automatically analyzes alerts using asset representation models before presenting them to maintenance teams. This intermediary provides pre-analyzed information, contextual relationships, and probable causes, enabling teams to investigate more efficiently without sacrificing thoroughness.
Solution Approach 2:
The system creates simplified copies or representations of complex asset states and alert contexts that are easier for maintenance teams to understand. By translating technical sensor data into asset-centric language with contextual relationships, the system reduces investigation time while maintaining thoroughness.
5Measurement precision
If expert knowledge is required to understand sensor relationships, then alert interpretation accuracy is improved, but system complexity increases
Solution Approach 1:
The system implements self-service through automated asset representation models that encode expert knowledge about sensor relationships and asset behavior. The model automatically interprets alerts and provides contextual information without requiring human experts to manually analyze complex sensor interrelationships.
Solution Approach 2:
The patent creates a virtual copy or model of the physical asset that captures expert knowledge about sensor relationships. This digital twin or asset representation serves as a knowledge repository that automatically provides interpretation guidance, reducing the need for human experts while maintaining interpretation accuracy.
6Reliability
If redundant maintenance procedures are implemented, then system reliability is improved, but production costs increase
Solution Approach 1:
The system applies partial monitoring and maintenance actions based on actual asset conditions rather than universal redundant procedures. By using asset representation to identify which specific assets or components need attention, the system eliminates unnecessary redundant maintenance while maintaining reliability through targeted interventions.
Solution Approach 2:
The system dynamically changes maintenance parameters and thresholds based on asset representation and real-time conditions. This allows the system to adjust maintenance intensity and frequency to match actual asset needs, reducing costs of redundant procedures while maintaining system reliability.
Data Source
AI summary
The present invention discloses methods and systems for problem-alert aggregation and identifying sub-optimal behavior. Methods include the steps of: providing data-driven alerts for an asset, wherein the data-driven alerts associate real-world data measured and/or detected from the asset, and wherein entities are physical objects and/or processes; providing an asset representation including interrelations between the objects, processes, and sensors associated with the entities of the asset; associating the data-driven alerts with the respective entities which are interrelated in the asset representation; aggregating the data-driven alerts into events, wherein the events are groupings of related data-driven alerts having related entities according to the asset representation; scoring each event into an event score, wherein the event score represents an event importance, an event urgency, an event relevance, and/or an event significance; and generating a selected subset of the events and respective event scores, wherein the selected subset is based on the event scores.