Virtual Sensors for Missing Industrial Machine Parameters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Industrial machines often lack appropriate sensors for certain process parameters, and historical measurement data may not be available, especially for machines at different development stages, leading to gaps in measurement data and challenging the accurate operation and monitoring of processes.
Innovation Solution
A neural network is trained using historical data from reference machines with appropriate sensors, employing transfer learning and unsupervised domain adaptation to provide parameter indicators for process parameters, effectively acting as a virtual sensor to emulate missing measurement data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sensors are installed on all process parameters of industrial machines, then measurement data availability is improved, but device complexity and cost increase
Solution Approach 1:
The patent creates virtual sensors that copy the functionality of physical sensors through neural networks. These virtual sensors generate measurement data for process parameters without requiring actual physical sensors, thereby maintaining measurement data availability while avoiding the complexity and cost of installing additional hardware sensors on every parameter.
Solution Approach 2:
The patent replaces the mechanical sensor system with an information-processing system based on neural networks. Instead of using physical sensors to directly measure parameters, the system uses trained neural networks that process existing measurement data to infer parameters that would otherwise require dedicated sensors, substituting mechanical measurement with computational inference.
2Measurement precision
If historical measurement data is collected from all reference machines, then training data quality is improved, but data management complexity increases
Solution Approach 1:
The patent extracts only the essential features and patterns from historical measurement data that are relevant for training virtual sensors. Rather than managing and processing all raw historical data from reference machines, the system identifies and extracts key measurement patterns and relationships, reducing data management complexity while maintaining training data quality.
3Adaptability or versatility
If operators manually measure process parameters, then measurement flexibility is improved, but human safety risks increase
Solution Approach 1:
The patent uses virtual sensors to copy the measurement functionality previously performed by human operators. The neural networks replicate the operator's ability to assess process parameters through data analysis, eliminating the need for operators to physically access hazardous areas while maintaining measurement flexibility through software-based adaptation.
4Measurement precision
If physical sensors are used for all parameters, then measurement accuracy is improved, but system cost and maintenance requirements increase
Solution Approach 1:
The patent makes the neural network-based virtual sensors universal, capable of providing measurement accuracy for multiple different process parameters using the same infrastructure. A single virtual sensor system can adapt to measure various parameters by loading different trained neural network models, eliminating the need for dedicated physical sensors for each parameter and reducing overall sensor quantity while maintaining measurement precision.
Data Source
AI summary
An industrial machine (123) may not have a sensor for a particular parameter, so that a computer uses a neural network (473) to virtualize the missing sensor. The computer trains the neural network (373) to provide a parameter indicator (Z′) of a further process parameter (173, z) for the industrial machine (123) with steps that comprise receiving measurement time-series with historical measurement data from reference machines, obtaining transformation rules by processing the time-series to feature series that are invariant to domain differences of the reference machines, transforming time-series by using the transformation rules, receiving a uni-variate time-series of the further process parameter (z), and training the neural network with features series at the input, and with the uni-variate time-series at the output.