Optimize Plant Performance Using Multivariate Data Analysis

This course will help you prepare for the certification exam and the exam fee is waived with this course.
Manufacturers are collecting more data than ever but face the challenge of deriving useful information from these large data sets. Part of the challenge is understanding the relationships amongst all the available variables (raw materials, process conditions etc.) and how these relationships impact process performance (yield, final quality etc.). Aspen ProMV finds the independent driving forces affecting process performance and effectively extracts actionable insights from available historical data. Learn how to use this desktop software on your historical process data to improve understanding of key process relationships.

Audience:

Engineers / scientists / statisticians responsible for process troubleshooting, control and optimization

Training Details

  • Course Id:

    PMV101

  • Duration:

    2 day(s)

  • CEUs Awarded:

    1.4

  • Level:

    Introductory

Benefits

  • Gain the practical skills and knowledge to use multi-block modelling to model your process.
  • Improve understanding of key process relationships
  • Identify key contributors to poor process performance
  • Troubleshoot and correct recurring process problems
  • Optimize process performance

Approach

  • Clear guidance on fundamental topics
  • Multivariate analysis workflows
  • Hands-on software labs
  • Experienced instructor-guided demonstration

Pre-requisites

Some familiarity with the Empirical Modeling is helpful but not essential.

Agenda

Introduction to multivariate data analysis 
  • Explore and discuss the multivariate analysis
  • Discuss the concept of latent variables 
  • Review relevance to Industrial problems 
Principal Component Analysis
  • Discuss the concept of Principal Component Analysis and the advantages to handle multivariate data
  • Apply PCA to a standard dataset and interpret results
  • Interpret residual plots to estimate the number of components needed
  • Workshop: Analyze a Regional Food Dataset Using PCA
  • Workshop: Analyze the Quality of a Polymer Product Using PCA
  • Workshop: Analyze a Silicon Wafer Process Using PCA
Partial Least Squares, or Projection to Latent Structures (PLS) Analysis
  • Discuss the advantages of Principal Component Regression and PLS over Multiple Linear Regression
  • Develop PLS models that explain variability in X and correlate to Y
  • Review how PLS can be implemented to solve various industrial problems
  • Workshop: Analyze the LDPE Polymer Process Using PLS
Historical Data Analysis and Multivariate Monitoring (MSPC)
  • Conduct Historical Data Analysis for a PCA model
  • Develop multivariate SPC charts to identify special causes of variation
  • Perform Multivariate Statistical Process Control for online data
  • Explore how MSPC can optimize a process by reviewing a case study
  • Workshop: Monitor the LDPE Process Using MSPC
Empirical Models Built from Historical Data
  • Review what constitutes a useful empirical model
  • Identify the differences between latent variable models and those developed from regression and machine learning methods
  • Discuss the limitations of empirical models developed from historical data
  • Workshop: Optimize the LDPE Process Using the Model Optimizer
Soft Sensors (Inferential Models)
  • Review the concept of soft sensors
  • Discuss the performance of these models very poorly
  • Workshop: Build a Dynamic PLS Model for the Kamyr Digester
Classification
  • Review the concept of classification
  • Discuss how PCA and PLS methods can be used effectively to classify data
  • Identify the industrial cases for classification
  • Workshop: Build PCA and PLS classification models on the Iris dataset

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.