Introduction to Process Troubleshooting
Step 1: Data Preparation
Data preparation is the first step required in order to prepare your data for modelling. Data is imported, and mathematical, statistical and dataset manipulation operations are applied to the data. Once the data source is properly configured, load the dataset into the Troubleshooter project, and continue with visualizing the data.
Step 2: Visualization
During this stage of the Process Troubleshooting methodology, data is visualized and prepared for modelling. Visual and statistical techniques are used to prepare the data that will be used in the following steps to build models. Thus it is important to use good quality data as well as data that is an accurate representation of the process. At this stage users should:
-
Set limits for process variables
-
Decorrelate the data set using a correlation matrix
-
Visualize and explore the data set using trends, histograms and scatter plots with the powerful history brushing capabilities to identify trends and clusters in the data set.
Step 3: Modelling
During the modeling stage of the Process Troubleshooting exercise, process models are constructed from historical data. Three types of models can be constructed from the prepared data:
Once a model is constructed, further views are provided on the model to extract new process knowledge from the data.
Step 4: Action Deployment
This step allows the models previously created to be exported to useful formats.
-
Export PCA
A PCA model can be exported as an XML file containing the eigen vectors, eigen values, normalisation parameters, mean values and field names contained in the PCA model. -
Export PLS
A PLS model can be exported as an XML file containing the inputs, output, factor loadings, x-coordinate weights, y-coordinate weights, regression weights and coefficients contained in the PLS model. -
Export Blueprint
The underlying Blueprint used in the Discrete & Batch Troubleshooter project can be exported for further editing in the Architect.
Related topics: