Futimo, fuzzy time-series data modeling method for modeling industrial time-series data.
This project contains scripts that allow data modeling using Futimo method. The method is implemented based on Tuomas Keski-Heikkilä's master's thesis. A journal article of the method is currently under writing.
For running this project Python 3 is required. Install the required packages by running
pip3 install -r requirements.txt
You can generate test data using generate_test_data.py. At the beginning of the script there are several parameters that can be used to modify the data. You can also use test_data_descriptions.txt to modify the gaussian functions used to generate data.
python3 generate_test_data.py
After that you can run fuzzy_modeling.py to fuzzify data and write it into database using the method desribed in [Placeholder for article].
python3 fuzzy_modeling.py
By default, this script also visualizes the fuzzy sets used for fuzzification.
In addition, it plots the raw data.
If you want to visualize data stored in database, run first expand_fuzzy_sets_for_visualization.py. This also aggregates data in database using the variables selected in the beginning of the file.
python3 expand_fuzzy_sets_for_visualization.py
Then you can run script data_visualization.py that produces visulization based on fuzzified data. This is illustarted below for the created test data.
python3 data_visualization.py
The variables to be visualized can be given as arguments:
python3 data_visualization.py -m "Temperature" "Motor speed rpm"
See Documentation for more accurate descriptions of the contents of this repository. In addition, if there are issues you can check Problems?.
We have collected a industrial crane dataset that is available in Zenodo. The more accurate description of tha data set is also available in Zenodo.
The root of dataset folder contains combined_csv.csv file that includes the most relevant variables from all cycles.
In addition, it is possible to examine each test cycle separately from Raw data -> CycleX. Each test cycle folder contain CycleX_export_dataframe.csv that includes the most relevant variables, and is good starting point for examinine data.
To model data in combined_csv.csv or CycleX_export_dataframe.csv files, the following modifications needs to be made to fuzzy_modeling.py:
Change the fuzzy set description file to fuzzy_sets_crane.txt
FILE_NAME_FUZZY_SETS = "examples/fuzzy_sets_crane.txt"
Then set input file name to the csv file you want to model (in this case, it is assumed that the file is located to the same folder as fuzzy_modeling.py file):
FILE_NAME_TEST_DATA = "combined_csv.csv"
Change csv delimiter to ,:
CSV_DELIMITER = ","
Change date format to OPC UA (note: the timestamps in the data are incorrect, as the OPC UA server was showing year 2103):
DATE_TYPE = Date_type.OPC_UA
You can change the table name in database:
TABLE_NAME = "crane_data"
Finally, change the number of variables to 13.
NUMBER_OF_VARIABLES = 13
Then run:
python3 fuzzy_modeling.py
If you want to visualize crane data, you need to run expand_fuzzy_sets_for_visualization.py.
From that file you can modify the variables that you want aggregate:
LIST_OF_AGGREGATED_VARIABLES = ["BridgePosition", "LoadTare", "HoistPosition", "TrolleyPosition"]
or then you can select all variables:
USE_ALL_VARIABLES = True
Remember also to change:
Fuzzy sets description file to fuzzy_sets_crane.txt
FILE_NAME_FUZZY_SETS = "examples/fuzzy_sets_crane.txt"
Table name to match tablename in database:
TABLE_NAME = "crane_data"
Finally, run:
python3 expand_fuzzy_sets_for_visualization.py
After this you can use data_visualizaton.py to visualize data. Remember to give the variables you want to visualize as arguments:
python3 data_visualization.py -m TrolleyPosition BridgePosition
After running this script using aggregation of all variables and combined_csv.csv as an input, you should get the following visualizations.
- Riku Ala-Laurinaho
- Julius Pesonen
The authors would like to thank Tuomas Keski-Heikkilä, Miika Valtonen and Juuso Autiosalo for their support. The authors would like to thank Joel Mattila for testing the scripts. The authors would like to express their sincere gratitude to Innovaatiosäätiö.
MIT
First working version is published.