Project Title: Autoclave Fault Detection

This project provides data exploration, manipulation, and clustering capabilities for a multi-variable execution dataset. It contains functionalities to explore the executions, group executions by their phase, and discover insights that might help improve the processes.

Libraries

• Pandas
• Numpy
• Logging
• Os
• Json
• Datetime
• Sklearn
• Collections
• Dtaidistance
• Plotly
• Matplotlib
• Seaborn

Use pip install <library_name> to install these libraries.

Class

This project has a single class called executions_analyzer. The class has various methods for calling different functionalities of the project.

Functions

• init(): Initializes the class by setting the directory paths to the data and other required configurations.
• load_phase_conf_json(): This function loads the phase configuration saved as JSON, and initialize the construct of each sequence config and adds it to a list of sequences config.
• logger_config(): Configures the logger.
• visualize_single_phase(): Plots the data of a single phase in plotly and matplotlib visualizations.
• build_samples_for_cluster(): Builds samples for executions clusters. It groups the executions and removes the executions with duration time=0.
• apply_cluster_algorithm(): Applies a clustering algorithm to the provided sequence's sample data (variables).
• apply_classification(): It takes two samples (list of points in the same sequence) and returns the DTW distance.
• calculate_phase_durations(): It calculates the duration of each phase in each sequence.
• phase_summary(): Produces a dataframe from the phase durations with summaries on means, standard deviations, etc.
• group_two_phases(name_a, name_b, criterion): This function groups two phases into a single phase, based on a criterion. The function can either exclude partial executions, the state of the variables, or both from the groupings.
• visualize_states_of_two_columns(column_a, column_b, point_id, name_phase): Produces a scatter plot that visualizes the data of the two variables(columns) by point_id and displays the results by each phase(column name). It also adds information about the average of each phase (column) in a vertical line across the chart.
• visualize_states_of_two_columns_all_points(column_a, column_b): Produces a scatter plot that shows the data of the two variables(columns) by point_id for all phases.
• visualize_variables_states(name_phase): Produces a scatter plot that shows the data of all variables in a phase. The plot has a time on the horizontal axis and the variable values on the vertical axis.

Conclusion

The executions_analyzer class is a valuable tool for exploring the executions, aggregating all the executions by a given phase. You can also build a clustering model, apply classification, produce visualizations, and calculate phase durations. All of these functionalities are useful for analyzing complex datasets. By following the instructions outlined in this README, you should be able to use the project with ease.