Fluid Property Measurement Using Mass Flow and Deep Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for measuring fluid properties, such as density, viscosity, and surface tension, particularly for high-temperature fluids like molten metals, are time-consuming due to the need for extensive computational fluid dynamics simulations.
Innovation Solution
A deep learning model is trained using computational fluid dynamics simulations to predict fluid properties based on mass flow rate data, reducing the time required for measurements by leveraging pre-constructed models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If computational fluid dynamics simulation is used to measure fluid properties, then measurement accuracy is improved, but measurement time increases significantly
Solution Approach 1:
The patent applies preliminary action by pre-training a deep learning model using computational fluid dynamics simulation data before actual measurement. The model is trained offline with simulated mass flow rate data corresponding to various physical property parameter sets, so that when actual measurement is needed, the pre-trained model can quickly predict fluid properties without requiring time-consuming real-time CFD simulations.
Solution Approach 2:
The patent uses copying by creating a deep learning model that replicates the behavior of the computational fluid dynamics simulation. Instead of running the complex CFD simulation during measurement, the system uses the trained deep learning model which has learned to copy the CFD simulation's output behavior, providing accurate predictions much faster.
2Productivity
If deep learning model is used for prediction, then measurement time is reduced, but model training complexity increases
Solution Approach 1:
The patent applies taking out by separating the model training process from the measurement process. The complex model training is performed once in advance using pre-generated simulation data, and then the trained model is deployed for rapid measurements. This extraction of the training phase allows the measurement phase to be simple and fast.
Solution Approach 2:
The patent uses preliminary action by completing the complex model training work before actual measurements are needed. The deep learning model is trained offline using a dataset of simulated mass flow rate data and corresponding physical property parameter sets, so that during measurement, only simple prediction operations are required.
3Device complexity
If traditional draining vessel method is used, then equipment simplicity is maintained, but measurement time increases
Solution Approach 1:
The patent applies merging by combining the traditional draining vessel method with a deep learning prediction system. The physical draining vessel setup remains simple, but it is integrated with a pre-trained deep learning model that provides rapid property predictions, thus maintaining equipment simplicity while eliminating time-consuming CFD calculations.
Solution Approach 2:
The patent replaces the mechanical/computational CFD simulation system with an intelligent deep learning model. Instead of running complex fluid dynamics calculations during measurement, the system uses the trained neural network to predict fluid properties based on measured mass flow rate data, substituting heavy computational mechanics with efficient machine learning inference.
Data Source
AI summary
A method of measuring physical properties includes a number of operations. Simulated mass flow rate data of training physical property parameter sets are generated by a computing unit based on the training physical property parameter sets and vessel shape information. Deep learning model is trained by a processor based on a training data set with an input feature vector of the simulated mass flow rate data and an output feature vector of the training physical property parameter sets. Measured fluid is received by a loader through an opening of a vessel of the vessel shape information. Weight accumulation data of weight of the loader is measured by a scale during a time period to obtain measured mass flow rate data. The measured mass flow rate data is inputted to the deep learning model to obtain measured physical property parameter set of the measured fluid.