The software on this site has been developed within CPACT for CPACT Member use and is available to full members of CPACT through the members only pages.
All the CPACT toolboxes, except for DoEMan, are compiled and do not need MATLAB to run. It is necessary though to download and install the Mathworks MCR (Mathworks Compiler Runtime) tool to allow the compiled toolboxes to run.
CPACT Software Toolboxes
Pre-Screen - Data Pre-Screening Toolbox
A Multivariate Statistical Data Pre-screening Toolbox (Pre-Screen) has been designed and developed for use by the practising process engineer who wishes to pre-process and pre-screen process data prior to multivariate data analysis, process data modelling or building predictive and inferential models. Currently all commercial data analysis packages address the univariate situation which is inappropriate in today’s industrial environment, where a large number of highly collinear and noisy process variables are collected for use in process modelling or performance monitoring where the initial data cleaning and conditioning task can consume up to 75% of the modelling time. A software toolkit has been developed specifically with the aim of focusing on the industrial needs for the initial data pre-screening of large data sets. The core feature of Pre-Screen is that it has been specifically developed to make the analysis of large data sets as automated as possible without taking away the need for engineering science understanding. The toolbox builds on top of the MATLAB programming environment, with powerful user interface procedures providing user friendly, mouse/menu driven software. The Main Features of Pre-Screen include: Data Tags, Load and Save facilities, Data Plotting, Normality (Univariate and Multivariate), Summary Statistics (Mean, standard deviations, covariance, correlations, skewness and kurtosis), Missing Data analysis and rectification, Spurious (outlier) data Elimination, Data Transformations, Data Filtering, Cross correlation, Data Transformations (Mathematical and Time shifting), Scatter Plots to Observe Possible Relationships, Loadings and Contribution plots, Histogram Plots, Normal Probability plots, Action tracking, and plot copying to WORD files.
MultiDAT - A Multivariate Statistical Modelling & Statistical Process Control Toolbox
The MULTIvariate Data Analysis Toolbox (MultiDAT) is designed and developed for use by the practising process engineer who wishes to analyse process data for plant data interpretation, feature detection and process modelling. The toolbox builds on top of the MATLAB programming environment, with powerful user interface procedures providing user friendly, mouse/menu driven software. The core feature of MultiDAT is that it has been specifically developed to analyse two data sets: a reference or nominal data set i.e. data collected when the process is in a state-of-statistical control and only common cause variation is present and a second data set which is either a validation set (i.e. it has not been used in the development of the model) or alternatively, a data set which contains faults, typical of those found on the process. The main features of MultiDAT include: editing functions, chronological log, plotting functionality; Residual Analysis, Contribution Analysis, Regression Coefficients, R2 and R2 –adjusted, Cross Validation Modelling Methodologies; Multiple Linear Regression (MLR) or Ordinary Least Squares (OLS), Stepwise Regression, Principal Component Regression (PCR), Projection to Latent Structures or Partial Least Squares modelling (PLS); Multivariate Statistical Process Control – MSPC (Process Performance Monitoring): PCA based monitoring, PLS based monitoring; Exporting data, graphs and pictures to any Windows Application. Data Summary
BatchDAT - A Multivariate Batch Process Statistical Process Control Toolbox
The Multivariate Statistical Modelling and Batch Statistical Process Control Toolbox (BatchDAT) has been designed and developed for use by the practising process engineer who wishes to analyse batch process data for plant data interpretation, feature detection and process modelling. The toolbox builds on top of the MATLAB programming environment, with powerful user interface procedures providing user friendly, mouse/menu driven software. The core feature of BatchDAT is that it has been specifically developed to analyse batch data comprising a combination of 3 data types: assays, process data and seed data. Each type of data for a batch is contained in a separate file, e.g. seed data for Batch 2 in seed2.xls, seed data for Batch 2 in seed2.txt, and process data for batch 1 stored in Process1.txt, etc… For a BatchDAT file to be properly defined, process, assays or seed files must exist and contain the data in a defined layout. Data file types supported are ASCII based TAB separated (.txt), comma separated (.csv) or Excel files (.xls). Batch analysis operations follows two routes: (i) through-batch observation level – (Wold et al - Method 1) and (ii) between-batch comparison level – (Nomikos & MacGregor - Method 2). The Main Features of BatchDAT include: Load and Save facilities, Data Plotting and Descriptive Statistics, Trend Plots, Histogram and Normal Probability Plots; Three-way batch data unfolding for: (i) through batch observation analysis and (ii) between batch analysis for comparisons between batches; Principal Component Analysis (PCA), Multi-way PCA and Multi-way PLS monitoring, Batch Loadings and Contribution plots, Action tracking, and Plot copying to WORD files.
DoEMan - Design of Experiments for Calibration Modelling
DoEMan is a software tool to simplify the use of experimental designs for choosing suitable pre-treatment methods for calibration problems. The immediate benefit of using experimental designs (DoE) to select pre-treatment methods for calibration problems is that choosing the best pre-treatment method becomes a structured and objective activity. DoEMan2 is re-worked version of the initial DoEMan software where the user interface has been improved and the code re-worked to simplify future extensions and modifications. For the user, DoEMan2 should make it straightforward to set up a design to test the performance of different pre-treatment methods, and interpret the results. Save options allow for sharing of designs and results among a number of users. DoEMan requires the MATLAB programming environment and the Eigenvector Research PLS Toolbox.
NewNet - A Neural Network Modelling and Model Validation Toolbox
A Neural network modelling toolbox (NewNet) provides a user friendly interface to the MATLAB Neural Networks toolbox and has been designed and developed for use by the practising process engineers who wish to build nonlinear models using industrial process data for process predictive modelling, inferential estimation and software sensors. The main features of NewNet include: Static and Dynamic Neural Networks, Building Network Models from Minimal Data, Network Stacking for Robustness, Model Validation. NewNet requires the MATLAB programming environment and the MATLAB Neural Network and Optimisation Toolboxes.