Fluid Measurement Device With Learning-Based Bubble Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional measurement devices struggle to accurately measure fluid parameters when foreign matter, such as bubbles, is mixed in the fluid, especially in fluids where parameters vary over time, leading to measurement errors.
Innovation Solution
A measurement device equipped with a detector and an arithmetic unit that performs regression and classification processes using a learning model built in advance, allowing it to accurately measure parameters even when foreign matter is present, by correcting for variations and anomalies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional measurement methods are used, then the device can measure fluid parameters, but measurement accuracy deteriorates when foreign matter such as bubbles is mixed in the fluid
Solution Approach 1:
The patent introduces a learning model as an intermediary between the detector and the parameter calculation. The learning model processes the sensor values and learns the relationship between sensor readings and actual parameters, enabling accurate measurement even when foreign matter is present. The arithmetic unit uses this learning model to correct measurement errors caused by bubbles or other contaminants.
Solution Approach 2:
The patent changes the measurement approach by using multiple sensor values and processing them through a learning model rather than direct calculation. The learning model adjusts and transforms the raw sensor data into accurate parameter values, effectively changing how the measurement is obtained to compensate for the presence of foreign matter.
2Measurement precision
If conventional measurement methods are used, then the device can measure fluid parameters, but measurement accuracy deteriorates when parameters vary over time due to operation
Solution Approach 1:
The patent performs preliminary action by building a learning model in advance that learns the relationship between sensor values and parameters under various operating conditions. This pre-trained model can then handle time-varying parameters accurately without requiring real-time adjustments, as it has already learned the patterns of variation during the training phase.
Solution Approach 2:
The learning model serves as an intermediary that bridges the gap between varying fluid parameters and stable measurement requirements. It processes the time-varying sensor data and outputs stable, accurate parameter values by applying the learned relationships regardless of parameter variations.
3Measurement precision
If a learning model is introduced to improve measurement accuracy with foreign matter, then measurement precision improves, but device complexity increases
Solution Approach 1:
The patent makes the arithmetic unit universal by equipping it with both traditional parameter calculation capabilities and learning model execution capabilities. This multi-functionality allows the same device to handle both conventional measurements and measurements with foreign matter present, without requiring separate dedicated systems.
Solution Approach 2:
The learning model enables the measurement device to self-adjust and self-correct for the presence of foreign matter. The device automatically applies the learned relationships to compensate for measurement errors, eliminating the need for manual calibration or external correction systems.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A measurement device 1 according to the present disclosure is a measurement device 1 configured to measure at least one type of parameter indicating the state of a fluid. The measurement device 1 includes a detector 10 configured to acquire, as data, at least one type of sensor value required for calculating the parameter, and an arithmetic unit 30 configured to calculate the parameter based on the data acquired using the detector 10. The arithmetic unit 30 is configured to acquire the result of at least one of a regression process or a classification process related to the parameter, which has been performed using the acquired data and a learning model built in advance based on the data when foreign matter has been mixed in the fluid.