Industrial AI: is it not that easy?
Manufacturers are traditionally a bit late in AI implementation compared to other businesses. By the way, the crisis, caused by the Covid-19 pandemic, has messed things up for many industrial giants on their way to transformation. However, AI provides a range of opportunities for neural networks application in industry. Some projects, that have already been implemented by some leading companies, confirm that.
So why is industry lagging behind in AI implementation? There is no simple answer to this question. On the one hand, big amounts of data are indispensable to assure neural networks’ successful work and many manufacturers are still using time-proven equipment without any sensors. That leads to lack of data on production processes. On the other hand, the cost of every step of production is considerably higher then, for instance, in commerce. That is why taking any modernization decision takes more time and requires a complex financial, standards, risks and security analysis.
Most of the industrial AI implementation projects, covered by the media today, were prepared several years ago and have already begun to yield results.
Production line automation
Real time monitoring turns out to be more efficient when performed by artificial intelligence rather than by an operator. Constant analysis of data received via sensors and cameras makes it possible to track production defects as well as to control the delivery terms.
Machine learning algorithms are capable of comparing data from many work shifts and locations, revealing hidden patterns and production bottlenecks. Thus, BMW company has been using AI for several years already in order to control the components on the production line in the process of the manufacturing. This lets the manufacturer identify any deviations to the standards even before the assembly, hence save time and money.
Better security
Artificial intelligence can not only supervise the production but also analyse people’s behavior on site. Video surveillance systems, that every factory has, can be easily transformed into video analytics tools equipped with automated incident alert systems.
Besides, computer vision may help operators of various machines. Thus, RZD (Russian Railways) has implemented neural networks as an additional instrument for coordinators, providing them with the most efficient schemes of rail wagons circulation at the shunting stations.
Additional “eyes” help avoid crashes and incidents and make shunting 20% more efficient.
Equipment and transport optimization
Neural networks are also able to control the vehicles’ circulation. Thus, Caterpillar company has been using AI technologies for the machinery in dangerous locations, such as mines and quarries. Neural networks analyse the real time data from sensors and detectors in order to identify obstacles and therefore avoid crashing into people and other machinery.
Russian Copper Company (RCC) has also been using AI for transport circulation optimization. Numerous detectors allow them to analyse the quality of roads, fuel consumption, itineraries and to check the weight of every vehicle transporting the ore. In these situations we are not talking about tonnes but about hundreds of tonnes so the company should know all the details about each vehicle in order to repair it, change the parts and gas it up on time.
A half an hour delay results in considerable transportation losses and profit decrease.
Assembly and manufacturing
Another way is to replace all the workforce with robots. By the way, many companies have started to consider this option during the pandemic, as working with robots, remote work as well, helps keep social distancing.
Thus, LG kitchen appliances factory, which required $525 million investment, is fully managed by a neural network. Every step of production and supply as well as final products quality control is held by AI. This makes the production highly effective
while self learning algorithms constantly improve the processes. The factory is expected to produce up to 3 million product units per year by 2023.
Production of materials
Optimization of metallurgy is even more challenging. But using AI is far more beneficial as well. Thus, AI, that determines the optimal amount of ferroalloys, has been used in the BOS (oxygen converter) unit at Magnitogorsk Iron and Steel Works. Ferroalloys are indispensable to achieve the required standard of the material.
Thanks to AI an operator gets real time recommendations about additives use.
This practice has led to reducing the usage of expensive ferroalloys by 5%. Therefore, AI is expected to save up to RUB 23 million per month.
Oil production optimization
Knowledge-intensive productions such as oil and gas production use artificial intelligence to optimize expensive mining operations. Thus, Gasprom Neft has a Cognitive geologist. Based on the neural network the company has been developing self learning models of geological units. That helps make more informed key decisions on mining exploration and answer the main questions about the profitability of the mineral extraction.
Manual evaluation of this data takes up to two years with the accuracy rate 60% max. Cognitive geologist allows companies to interpret geological data six times faster. And the quality of the information received is supposed to increase by 30%.
Back office processes
However, we shouldn’t forget about the fact that industries have a lot of processes which are not directly related to the production. Such as accounting, HR, logistics etc.
For instance, RZD has been using AI based systems for treating claims. The so-called “Electronic claims manager” automatically analyses claims from the natural monopoly clients and helps deal with thousands of requests faster and without additional personal contact and office meetings.
How to finance industrial AI?
Open Innovations forum, held online in 2020, brought forward various viewpoints on industrial AI. Thus, some companies have completely abandoned the idea of this expertise development, because some of their customers are facing market recession alongside the fall of demand and shortage of financing. That is why many manufactures, which keep on running digitization programs, do not include emerging technologies experiments in their 2021 budget.
At Smartcoders we stick to another point of view and offer our customers to develop the expertise together. There is absolutely no need to invest millions in some questionable experiments in order to implement AI, in industry as well as in any other fields. Instead you can choose a trusted and long lasting partnership when the professional company takes up AI implementation.
In this case the customer is able to benefit from AI usage without any investment at an early stage.
