Predictive maintenance convolutional neural networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current predictive maintenance techniques lack efficiency and effectiveness in utilizing telemetry data from monitored systems, such as heating, ventilation, and air-conditioning systems, to generate accurate maintenance predictions and interpretative metadata.
Innovation Solution
The development of predictive maintenance convolutional neural networks that process telemetry data to identify maintenance-critical time units and generate maintenance predictions, along with the use of per-sensor predictive input channels and heatmaps for explanatory metadata, enabling automated and interpretable predictive maintenance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional predictive maintenance techniques are used to process telemetry data, then the system can generate maintenance predictions, but the efficiency and effectiveness of utilizing telemetry data is insufficient
Solution Approach 1:
The patent replaces traditional mechanical predictive maintenance techniques with a neural network-based system. The neural network automatically learns patterns from telemetry data without requiring manual feature engineering or threshold setting, thereby improving both processing efficiency and prediction accuracy simultaneously.
Solution Approach 2:
The patent transforms the approach by changing from fixed threshold-based maintenance triggers to dynamic, learned parameters from neural networks. The system learns optimal maintenance timing parameters directly from historical telemetry data and maintenance records, enabling more accurate and efficient predictions.
2Measurement precision
If detailed telemetry data from all sensors is processed to improve maintenance prediction accuracy, then the prediction effectiveness improves, but the data processing complexity and computational resources increase
Solution Approach 1:
The patent segments the telemetry data processing by creating separate input channels for different sensor types (e.g., temperature, pressure, vibration). Each channel processes specific sensor data independently through the neural network, which simplifies the overall processing complexity while maintaining comprehensive analysis of all sensors.
Solution Approach 2:
The patent performs preliminary data processing and feature extraction within the neural network architecture itself. The network automatically pre-processes raw telemetry data, extracts relevant features, and identifies patterns, eliminating the need for complex external data processing pipelines.
3Speed
If maintenance predictions are generated without interpretable metadata, then the processing speed is faster, but the understandability and trustworthiness of predictions decrease
Solution Approach 1:
The patent introduces heatmaps as an intermediary visual representation that bridges the neural network's internal processing and the user's understanding. The heatmaps display which input data points and time periods most influenced the maintenance prediction, providing interpretable metadata without slowing down the core prediction generation process.
Data Source
AI summary
Systems and methods provide techniques for performing predictive maintenance data analysis based on telemetry data. In one embodiments, a method includes at least operations configured to obtain a training telemetry data object, determine a training input data object based on the training telemetry data object, obtain a maintenance data object, and generate a trained predictive maintenance convolutional neural network based on the training input data object and the maintenance data object. The trained predictive maintenance convolutional neural network can be utilized to generate maintenance predictions for a monitored system. The maintenance prediction can include data objects with visual explanatory capabilities, such as data objects that describe heatmaps over input telemetry data.