AI Process Control for Faster Steady-State Manufacturing Transitions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Continuous manufacturing processes face challenges in quickly transitioning from one output to another or maintaining output stability due to changes in inputs, such as feed stock properties or equipment conditions, leading to inefficiencies and waste.
Innovation Solution
A method and system utilizing machine-learning models, like artificial neural networks, to determine optimal input controls that minimize the time to reach a steady state by analyzing lag times and output parameters, and a system-on-a-chip (SoC) for real-time control adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional control methods are used to transition between outputs or maintain output stability, then the process can eventually reach a steady state, but the transition time is prolonged and efficiency is reduced
Solution Approach 1:
The machine learning model is trained in advance on historical process data to learn optimal control strategies for transitioning between steady states. During actual operation, the pre-trained model provides real-time control recommendations, eliminating the need for slow traditional trial-and-error adjustment methods and enabling rapid transitions.
Solution Approach 2:
The system continuously monitors process parameters during transitions and feeds this data back to the machine learning model. The model uses this real-time feedback to dynamically adjust control inputs, optimizing the transition path and accelerating the return to steady state while minimizing waste product generation.
2Loss of time
If multiple input parameters are adjusted simultaneously to accelerate transition, then steady state may be reached faster, but process control complexity and risk of instability increase
Solution Approach 1:
The machine learning model analyzes historical data to identify which input parameters have the most significant impact on transition dynamics. The system prioritizes adjusting these key parameters first while maintaining others at steady-state values, simplifying the control strategy while still achieving rapid transitions without excessive complexity.
3Loss of time
If aggressive control actions are taken to minimize transition time, then steady state is reached faster, but product quality may deteriorate during the transition period
Solution Approach 1:
The machine learning model determines the optimal degree of control aggressiveness by analyzing historical transitions. It applies just enough control action to accelerate the transition sufficiently while staying within bounds that prevent excessive degradation of product quality. The model balances transition speed with quality maintenance by leveraging patterns learned from past successful transitions.
Data Source
AI summary
Certain manufacturing processes, such as crude oil refining, operate continuously, wherein a facility produces one product as the feed stock for that product changes and/or the output product transitions from one product to another. Inputs to the process include physical items (e.g., feed stock) as well as control inputs (e.g., temperature, pressure, etc.), and often a change in one affects one or more others. As a result, a facility may take time to reach a steady state. An artificial intelligence is provided to model the facility and process and generate a second command input to reduce the time required for the facility to reduce the lag time until the facility has returned to a steady state.