AI and LPC

From May 2024, there is a new subsidy program for energy savings, and I expected an investment boom in heat treatment operations, aimed at this area to start. And since the majority of heat treatment takes place in furnaces with high energy consumption, I thought that a new period was coming with a targeted replacement of furnaces and technologies.

This has not happened so far. But since I represent ECM, I am slowly becoming an expert on the economics of carburizing under reduced pressure. The questions are still the same.

• Does LPC have any advantages?
• What do we actually save if we buy a furnace with LPC, while the majority of our business is in the field of hardening and tempering?
• We implement a large number of technologies on the existing equipment, not only carburizing, are these technologies transferable to a vacuum furnace equipped with LPC?
• And can we quench in a stream of nitrogen or do we have to continue with oil quenching?

And since I didn’t want to make up my mind, I asked the AI. Here is the answer. AI suggestions in italics, my comment in normal font. But sometimes I feel like I’m in Saturnin roman from Jirotka, an office for setting things straight.

My question on AI was: Carburizing under reduced pressure, can this technology replace classic multi-purpose furnaces, and how does it work out economically when I compare furnaces of exactly the same size?

AI answer:

If we compare vacuum carburizing and conventional atmospheric furnace carburizing of the same size, the differences focus more on operating characteristics, performance, surface quality and efficiency than on capacity alone. Here are the key factors to consider when comparing ovens of the same size:

1. Surface quality and accuracy:

• Vacuum furnace: Thanks to the vacuum environment, the carburizing process is cleaner and the surface of the parts has less oxidation or contamination. This process is particularly suitable for parts where surface finish and precision are important. The risk of inhomogeneities or uneven carbon deposition is minimized.
• Atmospheric furnace: During carburizing in the atmosphere, some oxidation of the surface occurs and a carbon layer may form, which requires subsequent processing (e.g. grinding or polishing), which is not necessary in vacuum carburizing.

[Source ECM]

JST: Yes, the AI ​​is not wrong here. LPC works in the absence of oxygen, so intergranular oxidation in this process is non-existent or minimal. Data according to ISO 15787:2016, IOD (Internal Oxidation Depth) in the case of LPC does not need to be prescribed. AI is also right about the cleanliness of the surface. Compared to classic carburizing, it is absolutely perfect.

2. Speed ​​and efficiency of the process:

• Vacuum Furnace: The vacuum carburizing process can be faster due to more uniform heat transfer in a low-pressure environment and more effective control of carbon diffusion. In addition, deformations and stresses are minimized, which shortens the time required for further processing.
•  Atmospheric furnace: The carburizing process can be slower, mainly due to the need to control the composition and flow of the atmosphere. The cycle time can be longer because more time needs to be spent on subsequent processes such as cooling or surface treatment.

JST: Yes, the rate of carbon transfer from the atmosphere and its subsequent diffusion into the steel is significantly faster with vacuum technology than with classic carburizing. Part of this contribution is due to the fact that we work in a clean environment, but part is also due to the fact that we can go to higher temperatures, which significantly accelerate diffusion. With conventional furnaces we are limited by a temperature of 950 °C, with LPC we are usually limited by a temperature of 1050 °C. But that’s only in theory. We must always keep in mind the size of the austenitic grain and its disproportionate growth.

[Source ECM]

The table shows the process design for a 1,500 kg charge, with a CHD requirement of 1.5 mm, and a comparison of the times for heating, carburizing, cooling below A_c1, heating to quenching temperature, and quenching. The existing carburizing process is carried out at a temperature of 920 °C, the newly proposed one at 960 °C. It is a 18CrNiMo7-6 material. Process time can be reduced by more than 30%. The second table contains the same entry, but for CHD 3.0 mm. Time saving is already 46%.

[Source ECM]

The consequence of such shortened process times is also a lower investment, because for the same volume of production, there will be a lower demand for capacity, and therefore the number of furnaces.

The argumentation of AI in classical carburizing is wrong, or completely pointless. Someone probably taught the AI ​​something wrong here.

As for deformations, yes, they can be lower with LPC, but only if we quench in the gas stream. If we quench in oil, the conditions are similar to classic carburizing, and in that case the deformations will also be similar.

3. Process control:

• Vacuum furnace: Vacuum furnaces allow very precise control of the amount of carbon, pressure and temperature. This leads to more homogeneous results and a higher quality of material processing. The process is less susceptible to disturbances caused by changes in atmospheric composition.
• Atmospheric Furnace: Classic multi-purpose furnaces use controlled atmospheres, but the precision of controlling carbon and other elements is less perfect than in a vacuum. In addition, the atmosphere may be affected by gas leaks or fluctuations in the gas supply.

JST: Even here, the AI didn’t pull out.. Today, with few exceptions, only acetylene is used for LPC. Earlier efforts to use other gases or mixtures were tied to valid patents. By using acetylene, the problem was simplified in that we only have 2 mass flow meters, for C2H2 and N2,  with the obligation to calibrate it once a year.

Acetylene dissolved in DMF is recommended. (DiMethylFormamide is an organic compound with the formula HCON(CH₃)₂ ) that dissolves acetylene well.

[Source ECM]

However, since we carburize under very low pressures, up to 10 mbar, it is difficult to measure its final effect in a vacuum furnace, i.e. the carbon potential, in this case the carbon partial pressure. Even if it is possible, for example, with a quadrupole mass spectrometer, or with the OPTIX gauge from the company GENCOA, in both cases it is very expensive equipment. Therefore, in the case of LPC, the computational simulation method is used, where we do not calculate Cp but the carbon flux related to the surface of ​​the parts in mg/cm2. In order for it to work properly in practice, we need to know the surface of ​​the batch. If we overdo the carbon flux, soot will form in our furnace. If the carbon flow is lower than the required flow for the given surface of ​​the batch, we will not achieve sufficient saturation of the surface. Every supplier of furnaces with LPC therefore has simulation software, where we simulate the process based on our parameters and batch surface, and it calculates the amount of acetylene flow for us on the mass flow meter.

In classic carburizing, there are 4 methods to control Cp. Oxygen probe, dew point measurement and infrared gas analyzer, foil method. However, according to CQI-9, it is mandatory to have a supervisory measurement of Cp. The optimal solution is with a supervising probe, where each probe is calibrated at an interval of 6 months, with a three-month overlap. ISO 20431 tells us that we have to verify the potential twice a week, CQI-9 then daily. Therefore, with classic carburizing, the requirements for Cp validation will be many times higher than with LPC. At LPC, we do not need qualified people for this.

With LPC, the accuracy of the process setup depends on the simulation software and our ability to accurately determine the batch surface. In classical carburizing, the accuracy of measurement depends on many parameters and measuring devices.

4. Operating costs:

• Vacuum furnace: Higher investment costs can be compensated by lower operating costs in the long term (lower gas consumption, less need to clean parts after the process).
• Atmospheric furnace: Lower purchase price but higher operating costs due to the cost of the atmosphere (e.g. gaseous hydrocarbons, nitrogen) and the maintenance of the furnace, such as cleaning from carbon deposits or oxides.

JST: Although AI is not wrong on this issue, I cannot completely agree with the interpretation. The higher investment costs will only be valid if I do not take into account the economic benefit of LPC. So faster diffusion, higher temperatures. If I need 3 furnaces for classic carburizing, when installing LPC it can be reduced to only two. A detailed economic analysis is therefore needed. Example in the following picture. For the range of production with predominant carburizing, 3 new Ipsen Atlas M furnaces are needed, but only two ECM Eco966 or Flex type furnaces with two heating chambers. The result is obvious, classic carburizing will be the most expensive investment solution.

The situation is clearer with higher operating and maintenance costs. LPC and vacuum are clearly better compared to classic technology. One of the examples is, for example, the costs of validating the furnaces in the above-mentioned report. So, 2 or 3 carburizing furnaces or chambers, 4 tempering furnaces and 1 washing machine. The most expensive is the existing condition, but that is understandable, these are 30-year-old furnaces. But if I take Ipsen Atlas M as a basis, then this maintenance operation can save me 40% of costs.

5. Power consumption:

• Vacuum furnace: Higher initial energy costs: Vacuum furnaces require energy to create and maintain the vacuum, which can be energy intensive, especially when starting and stabilizing the vacuum system. Lower operating energy consumption: Once vacuum is achieved, the furnace consumes less energy due to the absence of gaseous atmospheres and more efficient heat transfer. Heat transfer in a vacuum is controlled more efficiently, meaning less energy is lost to the environment. Vacuum furnaces also use controlled cooling (eg inert gas) which is more energy efficient than traditional atmospheric methods. Less cooling requirements: The inert gas cooling process is more efficient and less energy intensive than cooling in liquid media used in atmospheric furnaces.
•  Atmospheric furnace: Higher overall energy consumption: In order to maintain a controlled atmosphere (carbon, nitrogen, or other gases), it is necessary to continuously add gases and maintain their exact composition. This means that atmospheric furnaces often have a higher ongoing energy consumption. Higher heat losses: In atmospheric furnaces, there is higher heat loss because a stable atmosphere needs to be maintained and the furnace is not as well insulated from the environment as a vacuum furnace. This increases the energy consumption required to maintain the temperature and atmosphere

JST: Again, the AI ​​interprets it in its own way. I have no idea what he means by the initial higher energy costs. These will be essentially constant throughout the life of the furnace if it is properly maintained. But I can lean towards AI in the other statements. Quenching in gas will probably be less energy-intensive than quenching in oil. Not because the quenching turbine does not need energy, but because I have to heat the oil bath permanently for operational availability, and it will therefore have a large idling cost. Of course, gas consumption is also incomparable. If we need ENDO gas in the order of m3, with LPC we will move with both C2H2 and N2 significantly lower.

And we could continue like this. So, AI has its limits and gaps in education. How we ask questions is also important. But I realized that it’s more about telling the AI ​​what to say with our question, rather than it actually telling us what it thinks.

This year saw the implementation of AI legislation, Regulation 2024/1689 of the European Parliament. Here you need to see AI on two levels. The first level are programs of the ChatGBT type, a software product with the character of a closed database, having its own owner and source of information. This system is closed. The provider must ensure that it is not risky within the meaning of the directive, but it does not have to ensure that it is intelligent. Because it is closed, the provider is also responsible for all future changes leading to improvements in the level of intelligence. So, it is such a big educational encyclopaedia.

In contrast, open AI lies where self-learning functions can be applied, but these are not accessible to the public. This is because they cannot ensure that the self-learning capability does not take the system out of risk. It could happen that the AI ​​becomes an aggressive individual, or an individual for whom its lie is a basic tool.

What is available to us is therefore only a commercially applicable, fully controllable database of knowledge. And so, the only certainty from my attempt to interact with the AI ​​is that in general both I and the AI ​​think LPC is a good thing. But unlike AI, I don’t need anything more than common sense for that.

Jiří Stanislav

September 21,  2024