The first metallurgical ML hackathon from OMK IT. wait what??? / Hebrew

The first metallurgical ML hackathon from OMK IT. wait what??? / Hebrew

Greetings! Contact the United Metallurgical Company.

We have important news for all ML-specialists and those who are not indifferent to the topic II.

We will not surprise you much when we say that OMK IT has been using artificial intelligence in its work for a long time. But we haven’t had an LLM hackathon yet. It’s time to announce it!

What task are we solving?

In metallurgy, there is such a concept as assortment. Simply put, this is a list of characteristics of a metal product. Let’s assume: “wide-strip rolled steel (basic version), steel grade 09G2S, rolled steel thickness 25, strength class 325, category 12 according to GOST 19281-2014.”

Knowing what kind of product we are dealing with, you can make a number of decisions regarding further work with it. For example, to determine the list of necessary tests for conformity of the product to GOST. Or to mark and define the acceptance limits of these tests.

All indicators and events are strictly regulated by Normative and Technical Documentation (NTD), there is no special variability here.

At first glance, our task is very simple. We need an LLM, at the entrance of which we submit the characteristics of the assortment. And at the output we get a list of necessary tests, properties and permissible limits.

This task has maximum practical application. Using a similar LLM, it will be possible to automate the search process by NTD. And, probably, it will help to avoid errors related to the human factor.

And how now?

Now this task is solved by technologists and employees of VTK, who know GOSTs, if not by heart, then close to it. They collect the necessary information for all texts and plates and enter it into information systems. When something is forgotten, they take a printout of GOST on the table and check it.

Why LLM here

At first glance, it may seem that AI is redundant in this work. After all, the task is linear. Essentially, we need a directory search engine that logically looks like a series of input conditions. Information is required under these conditions.

There are two problems here.

First, we can have many options of what will come to us at the entrance. It is often impossible to write down this variability manually, you need some kind of automated system that can learn on its own.

Secondly, NTD (that is, a set of GOSTs) is digitized very heterogeneously. It is definitely available in the form of paper guides, but in electronic form, if you are lucky. It may not be text, but, say, a PDF file or a PNG image.

In this way, we need to recognize the characteristics of the assortment and the request, find the necessary DSTU, get the necessary information from them and give it to the output. Already drawn to artificial intelligence?

An example of the connection between the characteristics of the assortment and their mechanical properties

Only in fact there are more characteristics, and the speakers are on the right. And the Customers can also write something more sophisticated in the comments to the order by epistolary.

Examples of prompts and their answers

A few examples of what we want to get.

What can arrive at the entrance as part of the product

  • Steel grade;

  • Roll thickness;

  • Strength class (not all NTDs need it);

  • Rolled metal category (not all NTDs need it);

  • NTD (normative and technical documentation of products);

  • Additional information about the rental type and performance (in our examples, broadband rental).

We are ready to receive and accept certain conditions from participants. Let’s say if your model can only work with a standardized prompt where the order of parameters is important – OK, discuss. We have been working with GOSTs all our lives, for us forms and rules are not empty sound. We are ready to formulate the requirements correctly.

Examples of LLM prompts and correct answers

Request #1

List the list of tests for wide-strip rolled steel (basic version), steel grade 09G2S, rolled steel thickness 25, strength class 325, category 12 for GOST 19281-2014.

Respond

Temporal strength, yield strength, relative elongation, impact strength -40 (KCU), impact strength (KCU) fur. aging at room temperature.

The following requests are aimed at determining the limits of the mechanical property. A property can only be limited to min/max or max/min values.

Request #2

What are the limits for the temporary resistance test for wide-strip rolled steel (basic version), steel grade 09G2S, rolled steel thickness 25, strength class 325, category 12 for GOST 19281-2014?

Respond

The minimum boundary is 450 MPa.

Request #3

What are the limits for the temporary resistance test for wide-strip rolled steel, steel grade St3sp, rolled thickness 20, category 5 for GOST 14637-89?

Respond

The minimum boundary is 370 MPa. The maximum limit is 480 MPa.

What do we give in return?

So far, we have not been able to solve the problem on our own. That’s why we want to ask for the help of specialists who may be able to find some fresh and working solution. We do not expect that you will be able to give a completely finished model with a hundred percent hit on the prom. Rather, a hackathon will be the first step in a big work. But who knows? Will there suddenly be someone who will solve the task immediately?

In any case, we will choose three winners who will each receive their own prize.

  1. First place. Certificate for purchases in Ozone for 75 thousand rubles.

  2. Second place. A certificate for shopping in Ozone for 50,000 rubles.

  3. Third place. A certificate for shopping in Ozone for 25,000 rubles.

With people or companies that have achieved the greatest success in solving the task, OMK IT will be ready to continue cooperation on commercial terms.

We remind you that this is an extremely practical task. Such problems are faced not only by OMK IT, but by everyone connected with heavy metallurgy. And the resulting product, if successful, will have huge prospects.

Separately, we note that we do not expect participants to have deep knowledge in the field of heavy metallurgy. And we are not waiting for knowledge of GOSTs at all. The task is exclusively for the correct search for information and communication “standardized promt” – “extracting from the necessary GOSTs”.

The “necessary” GOSTs are given below:

https://docs.cntd.ru/document/1200113779

https://docs.cntd.ru/document/1200000119

How to participate

What is required to participate in the new hackathon? You need to do the following:

  1. Write “participation” in the comments to this article.

  2. You can also ask clarifying questions in the comments. However, we suggest that all those interested go to this one right away TG group. It has our experts, they will help with more specific questions, do not hesitate to ask questions!

  3. You need to upload your solutions via form (Using GitHub).

  4. The deadline for receiving answers is November 3.

  5. We will summarize the results in a separate post after November 11.

Dare and good luck)

Related posts