This dataset was originally sourced here: All OpenMV.net Datasets
What do you like about this case study?
This distillation tower case study is great because it allows the simple application of various JMP and JMP Pro platforms to analyse a dataset which has a few challenging characteristics. These challenging characteristics include:
- Seasonal variation
- Non-uniform sampling intervals
- Various outliers which vary in severity
- Data from numerous sensors
Because of these challenges this dataset shows simply how JMP has no problems tackling these difficult datasets. Additionally, due to the different features and quantity of data in this cases study, it is a widely applicable case study. This means that the dataset can be used to demonstrate various platforms within JMP and JMP Pro without needing to switch to other examples.
Particularly for the predictive modelling side of the demo this case study simple and clearly explains how bootstrap forest can be used, initially, to identify key predictors. As well as how the model screening platform can be used to quickly build and compare models that are created from various predictive modelling techniques such as decision trees and neural networks.
Finally, the dataset comes from a real industrial process, which gives us the opportunity to demonstrate that JMP can handle real data and therefore could be readily applied to the customer's particular area of concern.
Who is it for?
This case study is ideal for chemical and petrochemical engineers. Although since the study involves a distillation column it could be used to explain these concepts to most scientists.
How should I use it?
The dataset can be used in two key ways:
1. It can be used a exploratory data analysis (EDA) demo to show the powerful capabilities of platforms such as graph builder and predictor screening. Stanley Siranovich presented an excellent discovery summit by using the data set in this way which can be found here: Analysis of a Distillation Process Stream Using JMP® Pro 15 (2021-EU-45MP-766) - JMP User Community
2. By using the various scripts included with the data set (mainly the numbered scripts) this dataset can be used to highlight the robustness of the predictive modelling tools in JMP Pro. This involves identifying key predictors of yield in the data set, constructing various predictive models of the system and then using those models to predict future yields or to find optimal conditions.
3. No scripts are included in the dataset for this but since the dataset is sensor output from a process it can also be used to demonstrate the functional data explorer. This is a particularly good dataset for this because it is from a real industrial process and includes seasonal variation which can be dealt with by using the Fourier model fitting.
What platforms can we use?
- Distribution
- Graph Builder
- Predictor Screening
- Fit Model
- Functional Data Explorer
- Model Screening
- Various predictive modelling techniques including: Neural Networks, Bootstrap Forest, Decision trees and many more
