Statistical Techniques For Monitoring Industrial Processes

Extract Process Health Status From Data

What you'll learn

Gain proficiency in developing process monitoring solutions for complex industrial systems using popular statistical techniques

Learn how to implement automated fault diagnosis techniques for multivariate systems

Understand the pros and cons of different monitoring techniques for univariate and multivariate process systems

Work on industrial-scale case studies to consolidate the subject-matter understanding

Confidently build statistical process monitoring (SPM) solutions

Requirements

No Python programming and machine learning experience needed. The course covers everything that you need to know to implement the covered statistical process monitoring techniques.

Description

Welcome to your course on Statistical Techniques for Monitoring Industrial Processes where you will learn about the mainstream univariate and multivariate statistical techniques that have proven useful over the years for health monitoring of complex process plants. You will put the concepts learnt into practice using process industry-relevant datasets. Modern industrial plants are complex and therefore, it is a no-brainer that plant monitoring is an essential activity. Without exaggeration, it can be said that 24X7 monitoring of process performance and plant equipment health status, and forecast of impending failures are no longer a ‘nice to have’ but an absolute necessity! This course will equip you with the tools necessary to develop process monitoring solutions that includes both the fault detection (is the process or a signal behaving abnormally?) and fault diagnosis (which variables are behaving abnormally) components.Why study SPM (statistical process monitoring)?While artificial neural networks and deep learning grab most of the limelight now-a-days, classical statistical approaches are still are the bedrock of industrial process monitoring and enjoy immense popularity. Compared to neural network models, multivariate statistical techniques like PCA (principal component analysis) and PLS (partial least squares) are simpler to understand, more interpretable, and easier to develop and maintain; several successful stories.  and give you equal if not better performance than very complex models.What will you learn?In this course, you will get step-by-step guidance for developing industrial level solutions for statistical process monitoring. Emphasis is placed on conceptual understanding and practical implementations. Specifically, you will: learn about univariate SPM where you want to monitor a single process variable and multivariate SPM where you want to monitor multiple variables that interact with each otherin addition to covering the conceptual and implementation details, you will undertake several case-studies where you employ the learnt techniques on industrial-scale systems. You will work with data obtained from actual and/or simulated stirred tank reactors, catalytic cracking units, furnaces, chemical plants, polymer reactorsOutcome of the courseOnce you have mastered these techniques, you will be able to handle the monitoring needs of majority of the industrial processes. PrerequisitesWe will not assume any prior Python programming experience. Section 2 of the the course provides a quick introduction to Python programming and the development environment. Also, no prior machine learning experience is required.

Overview

Section 1: Course Introduction & Syllabus

Lecture 1 Course Instructor's Message

Lecture 2 Course Syllabus

Lecture 3 Introduction to Statistical Process Monitoring

Section 2: Python Installation and Basics

Lecture 4 Introduction to Python

Lecture 5 Python Language Basics

Lecture 6 Scientific Computing Package: NumPy

Lecture 7 Scientific Computing Package: Pandas

Lecture 8 Other Useful Python Packages: Quick Look

Section 3: Univariate Statistical Process Monitoring

Lecture 9 Introduction to Univariate SPM & Control Charts

Lecture 10 Shewhart Control Charts

Lecture 11 CUSUM Control Charts

Lecture 12 Case Study: Monitoring Airflow in an Aeration tank

Lecture 13 Judging Performance of Control Charts

Lecture 14 EWMA Control Charts

Lecture 15 Shortcomings of Classical Control Charts

Lecture 16 Exercise: Use Control Charts to Monitor Stirred Tank Reactor Operations

Section 4: PCA-based Multivariate Statistical Process Monitoring

Lecture 17 Introduction to Multivariate SPM

Lecture 18 Introduction to PCA

Lecture 19 PCA – Under the Hood

Lecture 20 PCA – An Industrial Case Study

Lecture 21 PCA – Fault Detection

Lecture 22 PCA – Fault Detection Implementation

Lecture 23 PCA – Fault Diagnosis

Lecture 24 PCA – Fault Diagnosis Implementation

Lecture 25 Exercise: Use PCA to Monitor Fluid Catalytic Crackers

Section 5: PLS-based Multivariate Statistical Process Monitoring

Lecture 26 Introduction to PLS

Lecture 27 PLS – Under the Hood

Lecture 28 PLS Modeling of LDPE Reactor

Lecture 29 PLS – Fault Detection

Lecture 30 PLS – Fault Diagnosis

Lecture 31 Exercise: Use PLS to Monitor Refinery Furnace Operations

Section 6: Deploying SPM Solutions

Lecture 32 Taking SPM Solutions to End-Users

Section 7: Course Conclusion

Lecture 33 Extensions of Classical PCA and PLS

Lecture 34 Concluding Remarks

Process data scientists who are looking to learn about statistical process monitoring,Students of chemical engineering, process systems engineering, and process data science,Process industry professionals (process engineers, reliability engineers, plant performance managers, etc.) who are interested in data science and interested in deploying automated monitoring tools for their processes