Process State Prediction from Partial Measurements Using ML Simulation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Industrial process control faces challenges in predicting process variables due to incomplete information available for simulation models, where existing methods struggle to accurately characterize the initial state of the process, leading to suboptimal control actions.

Innovation Solution

A computer-implemented method using a trained machine learning model to map process snapshot records to initial state records, which estimates hidden variables, and then simulates the process using a simulation model to predict future process variables, allowing for the identification of optimal control actions based on historical data and optimization objectives.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a simulation model is used to predict process variables, then the prediction accuracy is improved, but the method fails when not all initial state information is available

Engineering Contradiction:

Improveprediction accuracyVSAvoidapplicability with partial information

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

A machine learning model is introduced as an intermediary between the incomplete process snapshot and the simulation model. This ML model estimates the missing initial state variables that the simulation model requires, enabling the simulation to run successfully with partial information while maintaining prediction accuracy.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The machine learning model performs preliminary estimation of missing initial state variables before the simulation model is executed. This preliminary action completes the information set required by the simulation model, allowing the main prediction task to proceed without interruption.

Inventive Principle:

Principle #10Preliminary action

2Ease of operation

If process snapshot records with incomplete information are used, then the ease of data collection is improved, but the ability to uniquely identify initial state is lost

Engineering Contradiction:

Improvedata collection easeVSAvoidinitial state identification completeness

Core Design Contradiction:

Ease of operationVSLoss of information

Solution Approach 1:

The machine learning model acts as an intermediary that bridges the gap between incomplete process snapshots and the complete initial state representation needed for simulation. It infers missing information from available data, preserving both ease of data collection and information completeness.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The direct mechanical/physical measurement system is replaced by a data-driven machine learning system that can infer unmeasured variables from measured ones, substituting physical sensors with computational estimation for the missing state variables.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS20240126222A1Predicting Process Variables by Simulation Based on an Only Partially Measurable Initial State

Publication Date: 2024.04.18 ABB (SCHWEIZ) AG

AI summary

A method for predicting based on the state of an industrial process at a first point in time that is described by a process snapshot record with values of a first set of variables a value of at least one process variable of the industrial process at a second, later point in time, includes mapping using a machine learning model the process snapshot record to at least one initial state record; providing the initial state record to a simulation model; simulating using the simulation model the further development of the process; obtaining from the simulation model a final state record; and determining based on the final state record the sought value of the process variable at the second point in time.