Neural Network Control for Membrane Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Control of product stream compositions in membrane-based chemical separation processes is challenging due to slow process dynamics, variations in membrane performance, and the need to compensate for changes in feed stream composition and ambient conditions, which existing control systems like PID controllers struggle to manage effectively.
Innovation Solution
A control system and algorithm that utilizes membrane models to predict separation performance under varying operating conditions, allowing for intelligent adjustment of key process variables to maintain product stream compositions within desired ranges by calculating and communicating optimal operating conditions based on membrane performance parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If PID controllers are used to control product stream compositions, then the control system is simple to implement, but the control precision deteriorates due to slow process dynamics and measurement delays
Solution Approach 1:
The patent replaces traditional mechanical PID control systems with a neural network-based intelligent control system. The neural network learns optimal control strategies from historical process data and provides control recommendations that adapt to the slow dynamics of membrane separation processes, thereby improving composition control precision without requiring complex mechanical control mechanisms.
Solution Approach 2:
The patent introduces an intermediate neural network layer between the process measurements and the control actions. This neural network intermediary processes the slow composition measurements and predicts optimal control moves, bridging the gap between delayed measurements and the need for timely control adjustments in membrane separation processes.
2Measurement precision
If gas chromatography is used to measure product stream compositions, then measurement accuracy is high, but the measurement speed deteriorates due to slow and discontinuous analysis
Solution Approach 1:
The patent implements preliminary action by using the neural network to predict future composition trends based on current and historical data before actual composition deviations occur. This allows the control system to prepare control moves in advance, compensating for the slow measurement speed of gas chromatography and maintaining better composition control.
Solution Approach 2:
The patent creates a virtual copy of the physical process through the neural network model. This digital twin replicates the membrane separation process dynamics and allows real-time prediction of composition changes without waiting for actual gas chromatography measurements, effectively bypassing the measurement speed limitation while maintaining accuracy.
3Duration of action of stationary object
If membrane performance parameters are allowed to vary with use, then the membrane can operate for extended periods, but the separation performance deteriorates due to physical changes in the membrane
Solution Approach 1:
The patent applies dynamics by making the control system adaptive to changing membrane performance over time. The neural network continuously learns from process data and adjusts control strategies to compensate for membrane degradation, allowing the system to maintain optimal separation performance throughout the membrane's operational life rather than degrading along with the membrane.
Solution Approach 2:
The patent implements feedback mechanisms where the neural network continuously monitors process outcomes and uses this information to adjust control actions. This closed-loop feedback compensates for membrane performance deterioration over time, maintaining consistent separation performance even as the membrane undergoes physical changes during extended operation.
4Stability of the object's composition
If the control system responds to changes in feed stream composition and ambient conditions, then product composition stability is maintained, but the control complexity increases due to multiple varying parameters
Solution Approach 1:
The patent applies universality by designing a single neural network control system that handles multiple varying parameters simultaneously. The neural network processes feed composition changes, ambient condition variations, and membrane performance degradation through a unified control architecture, maintaining product composition stability without requiring separate control mechanisms for each disturbance.
Solution Approach 2:
The patent uses parameter changes by allowing the neural network to dynamically adjust control parameters based on varying operating conditions. The system learns optimal parameter adjustments for different scenarios (feed composition changes, ambient conditions, membrane aging) and applies appropriate parameter modifications to maintain product stability, simplifying the control approach compared to multiple dedicated control systems.
Data Source
AI summary
Provided herein are systems and methods for controlling a membrane-based separation process. The systems and methods can comprise, by one or more computing devices: receiving a product concentration, which product is produced by a process comprising a membrane separation, which process is operated at a first set of operating conditions; optionally calculating a membrane performance parameter based at least in part on the first set of operating conditions; calculating a second set of operating conditions based at least in part on the membrane performance parameter and a model of the process, such that operation of the process at the second set of operating conditions is expected to produce the product within a desired concentration range; and communicating the second set of operating conditions to the process.