Currently, most steelmakers are struggling to reduce lost production, which is costly and cripples profitability. The most common manifestations of steel production loss include weight variance, internal quality problems, surface parameter defects, mechanical properties, etc. What all of these losses have in common is that they are process-driven and cannot be addressed by replacing or maintaining machinery, or changing a specific set point. Instead, the real cause lies deep within the process itself, creating challenges for product manufacturing.
How to reduce steel factory loss?
Israeli artificial intelligence startup Seebo Interactive conducted an informal survey of hundreds of leading continuous process manufacturing companies, including dozens of executives at Steelmakers. The same question was raised – how to reduce plant losses? Incredibly, almost all of them articulated the same problem. In order to reduce production loss, the factory invested a lot of resources to bring in the best talents in the industry. These process experts or engineers often manage to reduce some losses, but at some point, they seem to encounter invisible obstacles. Regardless of the exact type of loss, the way the problem was dealt with, or how much money and resources were invested, there was a huge gap between what they were aiming for and what they were able to achieve, and in most cases, that gap was impossible. bridged.
Faced with this gap, Seebo has coined the term—complexity gap. After clarifying this common problem, Seebo set out to try to understand why this gap occurs: When performing traditional local root cause analysis, there is an inherent, huge blind spot, that is to say, an unknown black hole, the black hole is often much larger than the process specialists actually know about their processes.
The steelmaking process depends on human decision-making, but humans have their own limitations. In fact, the steel production line ultimately relies on human calculation and decision-making, which can also be said to be the crystallization of human talent, wisdom and experience, so of course there will be some blind spots.
Process experts and production teams make dozens of critical process-related decisions every day. These decisions can of course be enhanced, informed and effectively enforced by analytics platforms, measurement tools and more. But at the end of the day, these decisions are made and executed by humans.
There are inherent limitations here, as everyone approaches problems with their own biases and preconceived notions. This is natural, especially when dealing with something as complex as a steel production line, where humans cannot always consider all alternatives, and engineers and specialists have no choice but to use their past experience, and knowledge, In some cases, ad-hoc analysis is even performed on the basis of intuition.
Generally speaking, the root cause analysis of the steel production line is as follows
Process experts or engineers study the problem and propose a theory or a set of theories based on experience and intuition, then select several variables that they think are most likely to cause the problem, perform several mathematical or statistical calculations on these variables, and finally draw a set of conclusions.
This is where the blind spots are so obvious because the human mind cannot analyze every variable of data on the production line, especially all the complex interrelationships between each variable, and how these relationships themselves lead to the given problem. In fact, ad hoc analysis is limited to A handful of tabs.
This type of analysis usually works, but sometimes it doesn’t, and such ad-hoc analysis certainly doesn’t have a long shelf life because the production line is always changing. So even a decision or analysis that is correct today may not be correct tomorrow, or even a few hours later. It is difficult for humans to perform continuous, multivariate analysis of all data.
At the end of the day, process-driven losses are multifactorial, with sudden changes in conditions on the production line that simply cannot be discerned by the naked eye because each individual data variable is still within the allowable range when captured individually. Therefore, artificial intelligence will play a huge role.
Make the right decisions with artificial intelligence
The practice of many cases has proved that artificial intelligence, especially industrial artificial intelligence, will not replace human beings soon. In contrast, industrial artificial intelligence provides teams with insights that are difficult for humans to achieve, assists in making correct decisions, and improves performance.
Through conversations with hundreds of continuous process manufacturers, Seebo eventually identified three key criteria for an AI solution that, if met, could enable teams to achieve continuous process dominance.
One, revealing the hidden cause. The “unknowns” that manufacturing teams report most are hidden causes of inefficiencies and lost production that they aren’t even aware of. By uncovering these hidden causes, efficiency can be increased to new levels, greatly reducing losses in the process.
Second, continuous, scalable multivariate analysis of all data. Another clear gap in the current state of affairs is the ability to analyze all data continuously at all times, taking into account all the complex interrelationships between different points throughout the production line.
While humans cannot do this, AI can certainly, especially by using machine learning algorithms with supervision to understand patterns of behaviour that often lead to loss.
Finally, focus on the process. In steel manufacturing, as in other process manufacturing industries, the key to process is that it cannot be viewed in isolation. AI can provide humans with keen insights, but those insights will only be powerful if the algorithms understand the unique complexities of the entire process.
Without process expertise embedded in algorithms, AI will simply analyze data without a unique context and draw wrong or incomplete conclusions. Such techniques are known in practice as “automated root cause analysis”. If manual root cause analysis is what holds back steelmakers, automated root cause analysis is the solution.
Automated root cause analysis performs continuous, multivariate analysis of entire data sets and reveals hidden causes of production losses that process experts cannot find. Thanks to process-based artificial intelligence technology, this is already a reality. This technology embeds sophisticated machine learning algorithms with deep process expertise for each production line, enabling the algorithms not only to simply analyze the data but to understand each unique production process to correctly analyze it in context data.
Manufacturing teams can answer three key questions with this automated root cause analysis. Process experts or engineers should focus on:
1) Why did the loss occur?
2) How to prevent these losses in the future
The production team should focus on
3) When should actions be taken to prevent these losses
Using artificial intelligence to reduce KPI loss
Overcoming the limitations of human analysis, be it quality, rejects, yield or yield, and revealing hidden causes of production losses, is the core benefit of AI for steelmakers.
With clear, real-time insight into why losses occur, how to prevent them, and when to act, even an average team can produce extraordinary results.
Article Source: World Metal Bulletin “Using Artificial Intelligence to Reduce Steel Production Losses”