Neural Network Flow Signal Processing for Real-Time Fluid Property Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for monitoring fluid properties in chemical plants are limited by the difficulty in measuring local properties of fluids, relying solely on bulk property measurements which provide chaotic and meaningless data, making it challenging to monitor meaningful rheological properties in real time.
Innovation Solution
A method and device using an artificial neural network to process and monitor flow signals in real time, involving data acquisition from sensors, normalization, recurrent mapping to generate reservoir vectors, updating parameters based on output and label data, and applying these vectors to new data to acquire labels representing fluid properties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If bulk property measurements are used to monitor fluid properties, then measurement simplicity is improved, but measurement precision and meaningful information extraction deteriorate
Solution Approach 1:
The patent introduces bulk property measurements as an intermediary to indirectly infer local rheological properties. Sensors measure bulk properties (pressure, flow rate, conductivity) which then serve as input features for machine learning models that predict local rheological properties, bridging the gap between easily measurable bulk properties and difficult-to-measure local properties
Solution Approach 2:
The patent replaces direct mechanical/local measurement systems with a combination of bulk property sensors and computational models. Instead of using complex local measurement devices, the system uses simple bulk property measurements combined with machine learning algorithms to achieve accurate rheological property monitoring
2Measurement precision
If local properties of fluids are measured directly, then measurement precision is improved, but device complexity and difficulty of measurement increase
Solution Approach 1:
The patent uses bulk property measurements as intermediaries to indirectly determine local properties. Rather than placing sensors directly in the fluid flow to measure local properties, the system measures bulk properties upstream and uses computational models to infer local conditions, avoiding complex intrusive measurement devices
Solution Approach 2:
The patent creates a virtual model or digital twin of the fluid flow system using machine learning. This computational copy reproduces the behavior and properties of the actual fluid system, allowing local properties to be determined from bulk measurements through the virtual model without physically accessing local flow regions
3Measurement precision
If machine learning models are trained with more features and data, then prediction accuracy is improved, but training time and computational resources increase
Solution Approach 1:
The patent selects a specific subset of bulk property features (pressure, flow rate, conductivity, impedance) that are most relevant for predicting rheological properties, rather than using all possible measurement features. This partial selection achieves good prediction accuracy while reducing training complexity and time
Solution Approach 2:
The patent transforms raw sensor measurements into normalized features and applies various preprocessing transformations. By changing the parameter representation (normalization, feature engineering), the model achieves better accuracy with fewer training iterations and reduced computational requirements
Data Source
AI summary
The method for processing and monitoring a flow signal in real time in accordance with one exemplary embodiment of the present disclosure may include acquiring first physical data for a first fluid moving in a pipe through at least one sensor, normalizing the first physical data for the first fluid and performing recurrent mapping on the normalized first physical data to generate at least one reservoir vector of an artificial neural network, updating a parameter between the at least one reservoir vector and data output through the at least one reservoir vector based on the output data and first label data corresponding to the first physical data for the first fluid, and acquiring a second label of second physical data for a second fluid by applying the at least one reservoir vector to the second physical data based on acquiring the second physical data through the at least one sensor.