Location-Based Predictive Model Filtering for Asset Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current asset-monitoring systems lack the ability to distinguish between representative and non-representative operating data, leading to inaccurate predictive models due to unreliable data from locations such as repair shops or tunnels, where assets may generate skewed data.
Innovation Solution
An asset-monitoring system that maintains data on locations of interest where operating data is unreliable, allowing it to disregard such data when defining or executing predictive models, thereby maintaining model integrity by identifying clusters or specific areas where assets tend to output unreliable data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If operating data from all locations is used to define predictive models, then the quantity of data available for model training increases, but the accuracy and reliability of the predictive models deteriorates due to inclusion of non-representative data from locations such as repair shops and tunnels
Solution Approach 1:
The system segments the operating data based on asset location, separating representative data from non-representative data collected at locations such as repair shops, yards, and tunnels. By dividing the data into location-specific categories, the system can selectively use only representative data for predictive model definition, thereby maintaining model reliability while still utilizing available data resources.
Solution Approach 2:
The system applies local quality by assigning different qualities or weights to operating data depending on the asset's location. Data collected at representative locations is marked as high-quality for model training, while data from non-representative locations is either excluded or downweighted. This ensures that the predictive models are trained on high-quality data without completely discarding the volume advantage of comprehensive data collection.
2Adaptability or versatility
If operating data from non-representative locations is included in predictive models, then the coverage of data collection is improved, but the precision of model predictions deteriorates due to skewed data from locations like repair shops
Solution Approach 1:
The system performs preliminary action by pre-identifying and categorizing locations as representative or non-representative before the predictive model definition process. Location data is analyzed in advance to determine which areas produce reliable operating data. This preliminary classification allows the system to maintain comprehensive data collection coverage while ensuring that only pre-validated representative data is used for training predictive models, thus preserving prediction precision.
3Reliability
If the system tracks and excludes data from locations of interest, then the reliability of predictive models is improved, but the device complexity increases due to additional location tracking and data filtering mechanisms
Solution Approach 1:
The system introduces an intermediary component - a location tracking and classification module - that acts as a mediator between raw operating data and the predictive model definition process. This intermediary automatically tracks asset locations, compares them against a database of known non-representative locations, and filters or flags data accordingly. While this adds a layer of complexity, it automates the reliability improvement process and reduces the need for manual data quality assessment.
Solution Approach 2:
The system implements feedback mechanisms where the asset-monitoring system continuously monitors asset locations and provides feedback to the data filtering process. When an asset enters a non-representative location, the system automatically adjusts data collection and processing behavior. This feedback loop enables dynamic adaptation to location-based data quality issues without requiring complex manual intervention, balancing reliability improvement with manageable system complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Disclosed herein is a computer architecture and software that is configured to modify handling of predictive models by an asset-monitoring system based on a location of an asset. In accordance with example embodiments, the asset-monitoring system may maintain data indicative of a location of interest that represents a location in which operating data from assets should be disregarded. The asset-monitoring system may determine whether an asset is within the location of interest. If so, the asset-monitoring system may disregard operating data for the asset when handling a predictive model related to the operation of the asset.