Machine Learning in Python for Process and Equipment Condition Monitoring, and Predictive Maintenance: From Data to Process Insights

Part 1: Introduction and Fundamentals
Chapter 1: Machine Learning, Process and Equipment Condition Monitoring, and Predictive Maintenance
1.1 Process Industry and ML-based Plant Health Management
-- What are process faults and abnormalities
1.2 Plant Health Management (PHM) Workflow
1.3 ML Modeling Landscape for Plant Health Management
1.4 ML Model Development Workflow
1.5 ML-based Plant Health Management Solution Deployment
Chapter 2: The Scripting Environment
2.1 Introduction to Python
2.2 Introduction to Spyder and Jupyter
2.3 Python Language: Basics
2.4 Scientific Computing Packages: Basics
-- Numpy, Pandas, Sklearn
Chapter 3: Exploratory Data Analysis: Getting to Know Your Data Well
3.1 Why Exploratory Data Analysis Matters
3.2 Nonlinearity Assessment Techniques
3.3 Gaussianity Assessment Techniques
3.4 Dynamics Assessment Techniques
3.5 Multimode Distribution Assessment Techniques
3.6 Data Characteristics Investigation of Tennessee Eastman Process Dataset
Chapter 4: Machine Learning for Plant Health Management: Workflow and Best Practices
4.1 ML Model Development Workflow
4.2 Data Selection
4.3 Data Pre-processing
-- Handling data imbalance
4.4 Model Evaluation
4.5 Model Tuning
Part 2: Univariate Signal Monitoring
Chapter 5: Control Charts for Statistical Process Control
5.1 Control Charts: Simple and Time-tested Process Monitoring Tools
5.2 Shewhart Charts: An Introduction
5.3 CUSUM Charts: An Introduction
5.4 EWMA Charts: An Introduction
5.5 Case Study: Monitoring Air Flow in an Aeration Tank
5.6 Pitfalls of Univariate Control Charts and Alternative Solutions
Chapter 6: Process Fault Detection via Time Series Pattern Matching
6.1 Time Series Anomalies and Pattern Matching
6.2 Fault Detection via Historical Pattern Search
6.3 Fault Detection via Discord Discovery
Part 3: Multivariate Statistical Process Monitoring
Chapter 7: Multivariate Statistical Process Monitoring for Linear and Steady-State Processes: Part 1
7.1 PCA: An Introduction
7.2 Fault Detection via PCA: Polymer Manufacturing Case Study
-- Fault detection indices
7.3 Fault Isolation via Contribution Analysis for PCA
7.4 PLS: An Introduction
7.5 Fault Detection via PLS: Polyethylene Manufacturing Case Study
-- Fault detection indices
7.6 Fault Isolation via Contribution Analysis for PLS
Chapter 8: Multivariate Statistical Process Monitoring for Linear and Steady-State Processes: Part 2
8.1 ICA: An Introduction
-- Deciding the number of independent components
8.2 Fault Detection via ICA: Tennessee Eastman Process Case Study
-- Fault detection indices
8.3 FDA: An Introduction
8.4 Fault Classification via ICA: Tennessee Eastman Process Case Study
Chapter 9: Multivariate Statistical Process Monitoring for Linear and Dynamic Processes
9.1 Dynamic PCA: An Introduction
9.2 DPCA-based Fault Detection
9.3 Dynamic PLS: An Introduction
9.4 Canonical Variate Analysis (CVA): An Introduction
9.5 Process Monitoring of Tennessee Eastman Process via CVA
Chapter 10: Multivariate Statistical Process Monitoring for Nonlinear Processes
10.1 Kernel PCA: An Introduction
10.2 Fault Detection using Kernel PCA
10.3 Kernel PLS: An Introduction
10.4 Fault Detection using Kernel PLS
Chapter 11: Process Monitoring of Multimode Processes
11.1 Need and Methods for Specialized Handling of Multimode Processes
11.2 Multimode Semiconductor Manufacturing dataset
11.3 K-means Clustering: An Introduction
11.4 Gaussian Mixture Modeling: An Introduction
-- Deciding the number of clusters
11.5 Fault Detection via GMM: Semiconductor Manufacturing Case Study
Part 4: Classical Machine Learning Methods for Process Monitoring
Chapter 12: Support Vector Machines for Fault Detection 196
12.1 SVMs: An Introduction
-- Hard margin vs soft margin classification
12.2 The Kernel Trick for Nonlinear Data
-- Sklearn implementation of support vector classifier
12.3 SVDD: An Introduction
12.4 Fault Detection via SVDD: Semiconductor Manufacturing Case Study
Chapter 13: Decision Trees and Ensemble Learning for Fault Detection
13.1 Decision Trees: An Introduction
13.2 Random Forests: An Introduction
13.3 Fault Classification using Random Forests: Gas Boiler Case Study
13.4 Introduction to Ensemble Learning
-- Bagging
-- Boosting
13.5 Fault Classification using XGBoost: Gas Boiler Case Study
Chapter 14: Proximity-based Techniques for Fault Detection
14.1 KNN: An Introduction
-- Application for fault detection for metal-etch process
14.2 LOF: An Introduction
-- Application for fault detection for metal-etch process
14.3 Isolation Forest: An Introduction
-- Application for fault detection for metal-etch process
Part 5: Artificial Neural Networks for Process Monitoring
Chapter 15: Fault Detection & Diagnosis via Supervised Artificial Neural Networks Modeling
15.1 ANN: An Introduction
15.2 Process Modeling via FFNN: Combined Cycle Power Plant Case Study
15.3 RNN: An Introduction
15.4 ANN-based External Analysis for Fault Detection in a Debutanizer Column
15.5 Fault Classification using ANNs
Chapter 16: Fault Detection & Diagnosis via Unsupervised Artificial Neural Networks Modeling
16.1 Autoencoders: An Introduction
-- Dimensionality reduction via autoencoders
16.2 Process Monitoring using Autoencoders: FCCU Case Study
-- Fault diagnosis via contribution plot
16.3 Self-Organizing Maps: An Introduction
-- Evaluating SOM fit
16.4 Visualization of Semiconductor Dataset via SOM
16.5 Process Monitoring using SOM: Semiconductor Case Study
-- fault diagnosis via contribution plot
Part 6: Vibration-based Condition Monitoring
Chapter 17: Vibration-based Condition Monitoring: Signal Processing and Feature Extraction
17.1 Vibration: A Gentle Introduction
17.2 Vibration-based Condition Monitoring: Workflow
17.3 Vibration Signal Processing
-- Frequency domain analysis
-- Time-frequency domain analysis
17.4 Feature Extraction from Vibration Signals
-- Time domain features
-- Frequency domain features
-- Time-frequency domain features
Chapter 18: Vibration-based Condition Monitoring: Fault Detection & Diagnosis
18.1 VCM Workflow: Revisited
18.2 Classical VCM Approaches: A Quick Primer
18.3 Machine Learning-based VCM: Motor Fault Classification via SVM
Part 7: Predictive Maintenance
Chapter 19: Fault Prognosis: Concepts & Methodologies
19.1 Fault Prognosis: Introduction & Workflow
19.2 Machinery health Indicators: Introduction & Approaches
19.3 Health Indicator Construction Using Vibration Signals for a Wind Turbine
Chapter 20: Fault Prognosis: RUL Estimation
20.1 RUL: Revisited
20.2 Health Indicator-based RUL Estimation Strategies
-- Health degradation modeling
-- Trajectory similarity-based modeling
20.3 RUL Estimation via Degradation Modeling for a Wind Turbine
20.4 RUL Estimation via ANN-based Regression Modeling for a Gas Turbine

This book is designed to help readers quickly gain a working knowledge of machine learning-based techniques that are widely employed for building equipment condition monitoring, plantwide monitoring, and predictive maintenance solutions in process industry. The book covers a broad spectrum of techniques ranging from univariate control charts to deep learning-based prediction of remaining useful life. Consequently, the readers can leverage the concepts learned to build advanced solutions for fault detection, fault diagnosis, and fault prognosis. The application focused approach of the book is reader friendly and easily digestible to the practicing and aspiring process engineers and data scientists. Upon completion, readers will be able to confidently navigate the Prognostics and Health Management literature and make judicious selection of modeling approaches suitable for their problems.

This book has been divided into seven parts. Part 1 lays down the basic foundations of ML-assisted process and equipment condition monitoring, and predictive maintenance. Part 2 provides in-detail presentation of classical ML techniques for univariate signal monitoring. Different types of control charts and time-series pattern matching methodologies are discussed. Part 3 is focused on the widely popular multivariate statistical process monitoring (MSPM) techniques. Emphasis is paid to both the fault detection and fault isolation/diagnosis aspects. Part 4 covers the process monitoring applications of classical machine learning techniques such as k-NN, isolation forests, support vector machines, etc. These techniques come in handy for processes that cannot be satisfactorily handled via MSPM techniques. Part 5 navigates the world of artificial neural networks (ANN) and studies the different ANN structures that are commonly employed for fault detection and diagnosis in process industry. Part 6 focusses on vibration-based monitoring of rotating machinery and Part 7 deals with prognostic techniques for predictive maintenance applications.

Broadly, the book covers the following:

• Exploratory analysis of process data

• Best practices for process monitoring and predictive maintenance solutions

• Univariate monitoring via control charts and time series data mining

• Multivariate statistical process monitoring techniques (PCA, PLS, FDA, etc.)

• Machine learning and deep learning techniques to handle dynamic, nonlinear, and multimodal processes

• Fault detection and diagnosis of rotating machinery using vibration data

• Remaining useful life predictions for predictive maintenance