Use Case: Predicting Steel Plate Fault in Manufacturing with DataRobot

This post was originally part of the DataRobot Community. Visit now to browse discussions and ask questions about the DataRobot AI Platform, data science, and more.

What’s the problem?

Steel plates, whether used for general construction, industrial applications, or highly critical applications, must be of a structural quality to ensure safety and provide good formability and high strength. Every steel plate produced has some level of defects and faults. Normally, plates that don’t pass muster are scrapped, but not until the end of an expensive testing process. The time, money, and expertise spent on unusable product lowers the margins on each plate and adds complications to the manufacturing process.

The challenge and solution

Traditionally, each plate (or a representative sample) runs through a gauntlet of tests designed to reveal faults. The goal in terms of efficiency and cost is to accurately identify defective steel plates as early in the manufacturing process as possible, saving time and expense. Predictive modeling leverages data from that process, allowing you to target tests and identify faults in steel plates as early as possible. With good, accurate models, a plant can scrap bad plates earlier for higher efficiency.

The challenge, however, doesn’t end there. Plants make constant adjustments and improvements to their processes. If the predictive models aren’t updated with new data, they will quickly become outdated and suffer from a degradation in accuracy. It’s necessary to set up retraining processes that automatically identify the best model available based on new information.

For example…

Michelle is responsible for maintaining the quality steel plates her company produces. She knows the ins and outs of the production process, from the temperature to pressure ranges needed to ensure pristine results. Automated machine learning offers a new way to build upon her expertise and leverage the data her team collects. Instead of worrying about hiring a data scientist to support her team, she and the other engineers can take their data and build their own models. Those models can then predict which steel plates will have the highest likelihood of faults and also identify the reasons that may be driving those defective results.

In this example you will be using a dataset provided by Semeion, Research Center of Sciences of Communication, Via Sersale 117, 00128, Rome, Italy. 

Training data: https://s3.amazonaws.com/datarobot-use-case-datasets/steel_plates_fault_training.csv
Prediction data: https://s3.amazonaws.com/datarobot-use-case-datasets/steel_plates_fault_testing.csv
Data Dictionary: https://s3.amazonaws.com/datarobot-use-case-datasets/Steel+Plate+Fault+-+Data+Dictionary.pdf

About the author

Linda Haviland

Community Manager

Meet Linda Haviland