Processing Plant Control Optimization for Predictive Model Mismatch
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional optimization methods for manufacturing and processing control systems are limited by their industry-specific nature, poor portability, and inability to account for randomness in raw materials or unexpected process breakdowns, leading to suboptimal production efficiency and revenue loss in industries like oil and gas.
Innovation Solution
A computer-implemented method that selects optimization algorithms based on predictive model types, using mixed-integer linear programs for piece-wise linear functions and augmented Lagrangian methods for non-linear models to derive optimal set-points and flows, accounting for operational constraints and capturing complex relationships between inputs and outputs in a network of processing plants.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional optimization methods are used for manufacturing control systems, then implementation is straightforward with existing tools, but production efficiency is suboptimal and revenue is lost due to inability to account for randomness and complex relationships
Solution Approach 1:
The patent replaces conventional mechanical/optimization approaches with a neural network-based intelligent system. The neural network learns complex non-linear relationships between process variables and outcomes, substituting traditional mathematical optimization methods with a data-driven approach that can capture randomness and complex interactions in manufacturing processes, thereby improving production efficiency without requiring explicit mathematical models
Solution Approach 2:
The patent transforms the optimization problem by changing from direct optimization of process parameters to training a neural network on historical data. The system learns optimal parameter settings through pattern recognition in training data, allowing it to adapt to random variations and complex relationships without requiring explicit mathematical formulations, thus resolving the contradiction between simplicity and effectiveness
2Adaptability or versatility
If industry-specific optimization methods are used, then they are tailored to specific processes, but portability is poor and cannot be adapted to various industrial processes
Solution Approach 1:
The patent creates a universal optimization system using neural networks that can be applied across different industries and processes. The neural network architecture serves multiple functions: it can be trained on data from any manufacturing process, adapts to different process dynamics, and provides optimization recommendations universally applicable across diverse industrial contexts while maintaining reliability through data-driven learning
3Productivity
If conventional control systems are used, then they regulate sub-processes based on real-time data, but they cannot predict outcomes or optimize overall production efficiency
Solution Approach 1:
The patent implements preliminary action by training the neural network on historical process data before actual optimization is needed. The system learns from past outcomes and complex relationships in advance, building a knowledge base that enables it to predict future outcomes and recommend optimal actions before they are taken, thereby improving overall production efficiency without losing information about complex process relationships
Data Source
AI summary
Aspects of the invention include implemented method includes selecting an optimization algorithm for the control system of a processing plant based on whether the control system is guided by a linear-based predictive model or a non-linear-based predictive model, in which a gradient is available. Calculating set-point variables using the optimization algorithm. Predicting an output based on the calculated set-point variables. Comparing an actual output at the processing plant to the predicted output. Suspending a physical process at the processing plant in response to the actual output being a threshold value apart from the predicted output.