Steel Manufacturer Reduces Scrap Rates – and Costs – with AI | DataRobot AI Cloud

• Read original article here

Operating for more than a century, NIM Group has grown into one of the industry’s most technologically advanced carbon steel providers. And to keep its edge, the company’s data analytics team looks for every opportunity to improve decision-making, from the factory floor to executive offices.

Ben Dubois, Director of Data Analytics, NIM Group, envisioned using data to improve functions such as quoting, inventory management, and even machine settings to improve scrap rates. For the latter, operators have typically relied on operators’ knowledge and experience, resulting in inconsistency and making it challenging to ramp-up new operators.

“We knew there were areas of the company where we could use data to add value, whether it’s improving accuracy in our decision-making, or being able to automate some of our decision-making,” Dubois said.

NIM brought in DataRobot AI Cloud to automate predictive analytics and expand the team’s capacity to support the business.

“Other AI products were trying to solve a specific problem,” Dubois said. “What I like about DataRobot AI Cloud is the ability to use it in any way you can think up, whether it’s a normal regression-type problem, or forecasting, or for many different use cases.”

In a proof-of-value project, with the help of DataRobot University and DataRobot’s Customer-Facing Data Scientists, Dubois was able to develop an accurate model and begin realizing value quickly. Just as important, he could see how their data affected the results – helping him explain models to business stakeholders.

DataRobot Data Prep helps automate prepping the data while AutoML creates advanced models. APIs then automate productionalizing the results on the shop floor. Then, they easily monitor models in production.

Among several applications, the company applies the platform to predicting machine settings for processing steel. By introducing the correct settings into the machine from the start of the process, they generate less scrap, thus creating significant cost-savings.

With the AI Cloud platform’s application programming interfaces (APIs), they gather information about jobs in real-time, run them through a model, and then feed optimal settings back to the machines. Completing the feedback loop, the company tracks the actual settings used and the corresponding scrap rates to refine the model further.

“By giving operators a starting point, we shorten the trial-and-error period,” Dubois said. “We’re making more accurate predictions over time. Our model will keep getting better and better.”

AI-derived machine settings deliver two key benefits: less experienced operators ramp-up more quickly and make more informed decisions. Secondly, NIM is able to generate more steel that can be sold rather than end up in a scrap yard.

NIM also applied DataRobot AI Cloud to forecast demand for inventory to ensure they stock accordingly. For that, DataRobot Time Series allows them to find relationships between the demand for their steel and the industries they serve, such as agriculture and energy. By generating more accurate forecasts, NIM prevents lost sales and excess inventory, both of which are costly to the business.

“We look at factors within and outside the company so we’ll have the right inventory at the right time,” Dubois said. “One of the cool things about DataRobot and machine learning compared to just a normal time-series regression problem is being able to put a lot more features alongside your time series to make better forecasts.”

NIM has generated business value from DataRobot AI Cloud for years and only recently brought on a data scientist to help expand its efforts. By saving time across the process, Michael Green, Data Scientist at NIM Group, can spend more time with stakeholders to understand business problems and features.

“With DataRobot AI Cloud, we don’t have to worry about the minutiae of building every detail of one model,” Green said. “Instead of taking weeks or months to go from raw data to a deployed model, now we can do that in less than an hour.”

As a data scientist, Green gains satisfaction from helping solve business challenges throughout NIM Group.