Hybrid Sensor Modeling for Delayed Intrusive Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In industrial processes like petroleum, chemical, and food engineering, continuous monitoring of physical variables such as pressures, temperatures, and chemical components is hindered by intrusive sampling methods and time delays in obtaining sensor data, leading to reduced productivity and missed opportunities for process tuning.
Innovation Solution
A hybrid sensor system utilizing a hardware processor that receives upstream and downstream sensor data, determines time windows for lag analysis, and trains a machine learning model like a neural network to estimate target variable values in real-time, creating a causality relationship data structure to link sensor data and control set points in industrial processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If intrusive manual sampling and off-line lab analysis are used to monitor chemical components, then measurement precision is improved, but loss of time increases and productivity decreases
Solution Approach 1:
The system performs preliminary processing of sensor data using machine learning models to predict chemical component concentrations in real-time, before actual lab analysis is completed. This allows the control system to act on predicted values while awaiting confirmation from intrusive sampling, effectively reducing the time delay without sacrificing measurement precision.
Solution Approach 2:
A machine learning model serves as an intermediary between high-frequency non-intrusive sensor data and low-frequency intrusive lab analysis. The model learns the relationship between readily available sensor measurements and chemical component concentrations, generating real-time estimates that bridge the gap between the two measurement approaches.
2Measurement precision
If intrusive manual sampling is used for chemical analysis, then measurement precision is improved, but the number of samples that can be practically sampled decreases
Solution Approach 1:
The system creates a virtual copy of the chemical analysis function using machine learning models that replicate the behavior of intrusive sampling based on patterns learned from sensor data. This virtual sensor provides continuous chemical component estimates without requiring physical intrusion into the process flow, thereby increasing the effective number of samples that can be analyzed per unit time.
3Productivity
If high-frequency non-intrusive sampling is used, then productivity is improved and time delay is reduced, but measurement precision for chemical components deteriorates
Solution Approach 1:
The system merges high-frequency non-intrusive sensor data with low-frequency intrusive lab analysis results by training a machine learning model on combined datasets. The model learns to map the easily obtained high-frequency sensor measurements to the accurate but slow chemical component concentrations, producing real-time predictions that maintain the precision characteristics of intrusive sampling while achieving the speed of non-intrusive monitoring.
Data Source
AI summary
A hybrid sensor can be generated by training a machine learning model, such as a neural network, based on a training data set. The training data set can include a first time series of upstream sensor data having forward dependence to a target variable, a second time series of downstream sensor data having backward dependence to the target variable and a time series of measured target variable data associated with the target variable. The target variable has measuring frequency which is lower than the measuring frequencies associated with the upstream sensor data and the downstream sensor data. The hybrid sensor can estimate a value of the target variable at a given time, for example, during which no actual measured target variable value is available.