PDStoolkit 0.0.2

A Python package to facilitate building process data science solutions including process modeling, monitoring, fault diagnosis, etc.

PDStoolkit

Table of Contents

1. Project Description
2. Documentation & Tutorials
3. Package Contents
4. Installation
5. Usage

Description

The PDStoolkit (Process Data Science Toolkit) package has been created to provide easy-to-use modules to help quickly build data-based solutions for process systems such as those for process monitoring, modeling, fault diagnosis, system identification, etc. Current modules in the package are wrappers around pre-existing Sklearn's classes and provide several additional methods to facilitate a process data scientist's job. Details on these are provided in the following section. More modules relevant for process data science will be added over time.

Documentation and Tutorials

• PDStoolkit_Manual.pdf (in Github repository) provides some quick information on the algorithms implemented in the package
• Class documentations are provided in the 'docs' folder in Github (Source Code) repository
• Tutorials are provided in the 'tutorials' folder in Github (Source Code) repository
• The blog post (https://mlforpse.com/intro-to-pdstoolkit-python-package/) gives some perspective behind the motivation for development of PDStoolkit package
• Theoretical and conceptual details on specific algorithms can be found in our book (https://leanpub.com/machineLearningPSE)

Package Contents

The main modules in the package currently are:

• PDS_PCA: Principal Component analysis for Process Data Science

  • This class is a child of sklearn.decomposition.PCA class
  • The following additional methods are provided
    • computeMetrics: computes the monitoring indices (Q or SPE, T2) for the supplied data
    • computeThresholds: computes the thresholds / control limits for the monitoring indices from training data
    • draw_monitoring_charts: draws the monitoring charts for the training or test data
    • detect_abnormalities: detects if the observations are abnormal or normal samples
    • get_contributions: returns abnormality contributions for T2 and SPE for an observation sample

• PDS_PLS: Partial Least Squares regression for Process Data Science

  • This class is a child of sklearn.cross_decomposition.PLSRegression class
  • The following additional methods are provided
    • computeMetrics: computes the monitoring indices (SPEx, SPEy, T2) for the supplied data
    • computeThresholds: computes the thresholds / control limits for the monitoring indices from training data
    • draw_monitoring_charts: draws the monitoring charts for the training or test data
    • detect_abnormalities: detects if the observations are abnormal or normal samples

• PDS_DPCA: Dynamic Principal Component analysis for Process Data Science

  • This class is a child of sklearn.decomposition.PCA class
  • The following additional methods are provided
    • computeMetrics: computes the monitoring indices (Q or SPE, T2) for the supplied data
    • computeThresholds: computes the thresholds / control limits for the monitoring indices from training data
    • draw_monitoring_charts: draws the monitoring charts for the training or test data
    • detect_abnormalities: detects if the observations are abnormal or normal samples

• PDS_DPLS: Dynamic Partial Least Squares regression for Process Data Science

  • This class is a child of sklearn.cross_decomposition.PLSRegression class
  • The following additional methods are provided
    • computeMetrics: computes the monitoring indices (SPEx, SPEy, T2) for the supplied data
    • computeThresholds: computes the thresholds / control limits for the monitoring indices from training data
    • draw_monitoring_charts: draws the monitoring charts for the training or test data
    • detect_abnormalities: detects if the observations are abnormal or normal samples

• PDS_CVA: Canonical Variate Analysis for Process Data Science

  • This class is written from scratch
  • The following additional methods are provided
    • computeMetrics: computes the monitoring indices (Ts2, Te2, Q) for the supplied data
    • computeThresholds: computes the thresholds / control limits for the monitoring indices from training data
    • draw_monitoring_charts: draws the monitoring charts for the training or test data
    • detect_abnormalities: detects if the observations are abnormal or normal samples

Instalation

Installation from Pypi:

pip install PDStoolkit

Import modules

from PDStoolkit import PDS_PCA
from PDStoolkit import PDS_PLS

Usage

The following code builds a PCA-based process monitoirng model using PDS-PCA class and uses it for subsequent fault detectiona and fault diagnosis on test data. For details on data and results, see the ProcessMonitoring_PCA notebook in the tutorials folder.

# imports
from PDStoolkit import PDS_PCA

# fit PDS_PCA model
pca = PDS_PCA()
pca.fit(data_train_normal, autoFindNLatents=True)

T2_train, SPE_train = pca.computeMetrics(data_train_normal, isTrainingData=True)
T2_CL, SPE_CL = pca.computeThresholds(method='statistical', alpha=0.01)
pca.draw_monitoring_charts(title='training data')

# fault detectiona and fault diagnosis on test data
pca.detect_abnormalities(data_test_normal, title='test data')
T2_contri, SPE_contri = pca.get_contributions(data_test_normal)

License

All code is provided under a BSD 3-clause license. See LICENSE file for more information.