Deep Learning Pattern Detection for Machine Behavior Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for identifying machine behaviors in heavy machinery and construction vehicles are limited by their reliance on manual correlation of sensor readings, which is time-consuming, expensive, and provides a narrow understanding of machine behavior, and requires prior knowledge of machine features.
Innovation Solution
A system utilizing deep learning algorithms to identify patterns in signals from machine sensors, combined with targeted testing to validate and understand the physical meaning of these patterns, enabling a broader understanding of machine operating states and behaviors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If manual correlation of sensor readings is used to identify machine behaviors, then the method provides interpretability and requires prior knowledge of machine features, but it is time-consuming, expensive, and provides a narrow understanding of machine behavior
Solution Approach 1:
The patent segments the machine behavior identification process into two distinct phases: (1) automated pattern discovery using deep learning on fleet-wide sensor data, and (2) targeted physical testing to validate and interpret specific patterns. This segmentation allows the system to leverage automated methods for comprehensive pattern detection while using manual methods only for critical validation, thus reducing overall time and cost while maintaining interpretability.
Solution Approach 2:
The patent introduces targeted physical testing as an intermediary between automated pattern detection and final behavior identification. This intermediary step validates the patterns discovered by deep learning algorithms and provides physical meaning to the detected behaviors, bridging the gap between automated data analysis and human-understandable machine behavior interpretation.
2Productivity
If deep learning algorithms are applied to identify patterns in sensor signals, then the system achieves automated and comprehensive behavior detection, but it requires validation to understand the physical meaning of identified patterns
Solution Approach 1:
The patent performs preliminary action by conducting targeted physical testing during the pattern validation phase to establish the physical meaning of deep learning-identified patterns before deploying the system for automated behavior identification. This preliminary validation ensures that the complex patterns detected by deep learning algorithms correspond to actual machine behaviors, reducing the complexity of subsequent interpretation.
3Measurement precision
If targeted tests are performed to validate deep learning patterns, then the system achieves accurate behavior identification with physical meaning, but it requires additional time and resources for testing
Solution Approach 1:
The patent applies partial action by performing targeted physical testing only for specific patterns that require validation, rather than testing all patterns discovered by deep learning. This selective approach maintains high measurement precision for critical behaviors while minimizing the time and resources required for validation.
Data Source
AI summary
Systems and methods are described for identifying a behavior of a machine. A computer system receives a signal indicative of operation of a field machine and applies a deep learning algorithm to identify a pattern in a collection of signals stored on a computer-readable memory. The collection of signals includes the received signal indicative of operation of the field machine and other signals. A series of targeted tests are performed using a test machine while monitoring a signal indicative of operation of the test machine. A behavior is identified during the series of targeted tests that produces a signal that matches the pattern identified by the deep learning algorithm. An occurrence of the behavior is then automatically identified in the field machine in response to detecting the pattern in the received signal indicative of operation of the field machine.