Predictive OEE Modeling Using Spectral Sensor Degradation Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining overall equipment effectiveness (OEE) in industrial equipment lack precision and fail to provide a robust predictive metric based on historical and real-time sensor data, especially in visualizing extended periods of equipment operation.
Innovation Solution
A predictive model using machine learning models that obtain spectral features from sensor data to determine a probability of survival, which is then used to calculate an overall equipment effectiveness metric as a product of predicted planned production time, performance, and quality output, enabling precise, high-resolution visualization of equipment productivity over time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional OEE determination methods are used, then the calculation process is simple, but the precision and robustness of the predictive metric is insufficient
Solution Approach 1:
The patent segments the OEE calculation into three distinct predictive components (availability, performance, quality) each modeled separately using machine learning, then combines them multiplicatively to form the overall predictive OEE metric. This segmentation allows each component to be optimized independently while maintaining overall system precision.
Solution Approach 2:
The patent transitions from traditional time-based OEE tracking to a spectral frequency domain analysis using Fourier transforms. By converting sensor data into spectral features and analyzing equipment degradation in the frequency domain, the system achieves higher measurement precision for detecting equipment states and predicting OEE metrics.
2Measurement precision
If spectral analysis of sensor data is implemented, then the visualization resolution and detail is improved, but the data processing complexity increases
Solution Approach 1:
The patent performs preliminary spectral analysis and feature extraction on sensor data before feeding it into the machine learning models. By pre-processing the raw sensor data through Fourier transforms and extracting relevant spectral features in advance, the system reduces the complexity of subsequent processing steps while maintaining high detection precision.
Solution Approach 2:
The patent introduces spectral features as an intermediary representation between raw sensor data and the OEE predictive models. These spectral features act as a mediator that captures essential equipment state information in a compressed, interpretable form, reducing the dimensional complexity while preserving critical diagnostic information.
3Reliability
If machine learning models are used for prediction, then the predictive accuracy for future OEE is improved, but the computational resources required increase
Solution Approach 1:
The patent implements a hybrid approach where only critical equipment parameters and spectral features are fed into the machine learning models, rather than processing complete raw sensor datasets. This partial action approach maintains predictive reliability by focusing on the most informative features while significantly reducing computational energy consumption.
Solution Approach 2:
The patent transforms raw sensor data into spectral domain parameters through Fourier transforms, changing the representation parameters to frequency components. This parameter transformation enables the machine learning models to work with more compact, informative features that require less computational power while maintaining or improving predictive accuracy.
Data Source
AI summary
Among other things, systems and techniques are described for a predictive model for determining overall equipment effectiveness (OEE) in industrial equipment. Data including spectral features is obtained. A probability of survival is determined by fitting at least one degradation function to degradation data associated with the industrial equipment. An overall equipment effectiveness metric is predicted as a product of predicted planned production time, predicted performance, and predicted quality output by trained machine learning models.