Inline Fluid Quality Monitoring with Neural Network Deviation Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In-line monitoring of fluid quality is challenging due to the difficulty in detecting changes immediately, especially in complex systems, where a single measurement value is insufficient to assess potential hazards, and traditional methods require lengthy sampling and pretreatment, limiting detection accuracy and frequency.
Innovation Solution
A method using a neural network to evaluate combined in-line measured values from multiple variables, projecting them into a reduced vector space, and employing a kernel density estimator to assess deviations from a reference state, allowing for rapid and reliable detection of fluid quality changes without requiring specific threshold values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If in-line measuring devices are used to detect fluid quality changes immediately, then detection speed and measuring frequency are improved, but the ability to achieve high detection accuracy and reliability is worsened
Solution Approach 1:
The patent combines multiple in-line measuring devices that measure different variables (e.g., turbidity, conductivity, pH, temperature) into an integrated monitoring system. By merging the data from these multiple sensors and applying multivariate data analysis, the system achieves both immediate detection speed and high reliability through the combined information from various measurement variables, resolving the contradiction between fast detection and accurate assessment.
2Reliability
If multiple in-line measured values with respect to different measurement variables are captured to assess fluid state, then detection reliability is improved, but system complexity and data evaluation difficulty are worsened
Solution Approach 1:
The patent introduces multivariate data analysis methods (such as principal component analysis, neural networks, or other statistical evaluation methods) as an intermediary between the multiple sensors and the final assessment. This intermediary processing layer automatically evaluates the complex multivariate data, reducing the perceived system complexity while maintaining high detection reliability through comprehensive analysis of all measurement variables.
3Measurement precision
If traditional sampling and pretreatment methods are used to achieve high detection accuracy, then measurement precision is improved, but measuring time and productivity are worsened
Solution Approach 1:
The patent replaces the mechanical sampling and pretreatment system with an in-line measurement system that directly measures fluid quality parameters in the process stream. By using multiple in-line sensors and multivariate data analysis, the system achieves detection accuracy comparable to or exceeding traditional methods while eliminating the time-consuming sampling and pretreatment steps, thus resolving the contradiction between precision and speed.
Data Source
AI summary
A method for automated in-line detection of deviations of an actual state of a fluid from a reference state is disclosed wherein measured values captured at the same time are evaluated in a combined manner with respect to at least three measurement variables that are different measurement quantities of the fluid and/or a measurement quantity of the fluid measured at different measuring points. The method includes creating a reference data set, wherein reference measured values are mapped to a reference vector of a vector space using a neural network; in-line measurement, wherein measured values at all times are mapped to a measurement vector using the neural network; comparing the measurement vector with the reference vectors using a kernel density estimator of a predefinable window width; and creating an assessment with respect to a deviation of the actual state from the reference state on the basis of the kernel density estimator.