Improve Model Accuracy using First Principles Driven Hybrid Models

Understand the advantages of using First Principles Driven Hybrid Models. Develop accurate hybrid models using both mechanistic and AI/ML fundamentals. Build and deploy predictive hybrid models within process simulators to improve accuracy for several units such as distillation columns, reactors, heat exchangers, pressure changers, separators, etc. Use plant data to replace inadequately modeled relationships not fully captured by traditional engineering models.

Audience:

Process Engineers

Training Details

  • Course Id:

    EHM024

  • Duration:

    1 day(s)

  • CEUs Awarded:

    0.7

  • Level:

    Introductory

Benefits

Leverage plant data to enhance first-principles models using AI/ML to improve modeling accuracy

Approach

  • Clear guidance on fundamental topics
  • Industry workflows
  • Hands-on workshops
  • Experienced instructor-guided demonstrations
  • Q&A on student-specific problems

Pre-requisites

Basic knowledge of Aspen Plus/HYSYS recommended

Subsequent Courses

EAP101, EHY101

Agenda

Introduction to First Principles Driven Hybrid Models

- Learn the general concepts and definitions of First Principles Driven Hybrid Models 

- Identify different types of workflows and their uses

- Recognize First Principles Driven Hybrid Modeling architecture and use cases


 Data Formatting and Pre-Processing

- Become familiar with the format required for building First Principles Driven Hybrid Models from plant data

- Identify different types of data pre-processing available

- Understand the new AI training interface in the simulator

- Learn how to import raw data into the simulator

Workshop 1: Format and Import Plant Data 


Analyzing and Conditioning Raw Data 

- Understand the tools to analyze raw data

- Identify trends and correlations within the data.

- Learn how to apply different conditioning techniques to raw data

Workshop 2: Analyze and Condition Raw Data


Building the Hybrid Model

- Evaluate the need to train a hybrid model for your process

- Build a hybrid model from conditioned plant data

- Select dependent and independent variables to be used in the hybrid model

- Identify a Neural Network Output to be trained in the hybrid model

Workshop 3: Evaluate and Train a Hybrid Model


Validating the Hybrid Model

- Identify best practices to validate the hybrid model

- Use the snapshot feature to try different data conditioning and NN configuration

- Understand key parameters to evaluate the accuracy of the hybrid model before its deployment

Workshop 4: Validating the Hybrid Model


Deploying the Hybrid Model

- Learn how to deploy hybrid models in the process simulator

- Explore the automatic changes made to the simulation interface once the model has been deployed.

Workshop 5: Deploying the Hybrid Model

Using and Sustaining the Hybrid Model

- Enter the minimum input required for running the hybrid model

- Recognize the usability of the model

- Identify the need to re-train the model with newly updated data

Workshop 6: Using and Sustaining the Model

Register for a Class

Date Class Type Location Price Language

Aspen Technology, Inc. awards Continuing Education Units (CEUs) for training classes conducted by our organization. One CEU is granted for every 10 hours of class participation.