The subject of business digitalization is on everyone's lips now. In some industries, technology is changing at an explosive pace. But what about petrochemistry?
In petrochemistry, such an explosive development is not observed and is not expected. Why not? Petrochemicals, if we consider its production area, has always had robust, automated processes involving process control systems. As for the Internet of Things, it came to the industry in the 70s, and it was used quite widely in the 80s. The same applies to business processes; so I would say that digitalization is going quite smoothly in our industry.
But in 2017, Sibur announced a transition to large-scale digitalization. What was meant by this?
We started working on a comprehensive digital strategy, which became one of management's top focuses, and we announced it in December 2017 at a corporate forum. There were previously many interesting digital projects in various areas, and in the spring of 2017 we decided that it was important not to miss the chance to introduce new technologies and to put together our own expertise and digital culture, so we began investing in individual projects and testing specific technologies.
At the beginning of 2018, we established two new divisions (Digital Technologies and Corporate Data Management) and began recruiting employees. Attracting digital professionals to the petrochemical industry proved to be challenging. They were familiar with banks and retail, but in this case, it was as if we were offering a pilot the chance to become a Formula 1 driver. Now, almost 200 people are involved in these divisions, including developers, UI/UX designers, data scientists, data analysts, and others. Before that, there were no such professions in Sibur.
What did it give you?
Generally speaking, we have accelerated the process of adopting new digital solutions. For example, there are 2–2.5 times more projects in our pipeline than a year ago. Internal functional customers (from logistics specialists to procurement managers) began to develop more actively in terms of innovation.
The second important point is that we didn't have big data analysts. Previously, we worked with contractors to implement such projects, but we did not have our own internal expertise. Now we have it, and last year we independently implemented such projects.
Third, we now have focused significantly more human resources and efforts (i.e. projects) on using different devices (wearables or sensors). Moreover, the next wave of sensors is coming, and they are getting cheaper.
What is the company's IT management structure now?
As I said, we started our own development, and this in itself is a large and complex project, since programmers need an environment for developing, testing, and DevOps.
When the company acquires in-house development, system changes can be made much faster. Do you remember that Sberbank's German Gref complained that it took too long for them to implement changes? We have also encountered this difficulty, but we understand how to handle it. There are three factors: how quickly we decide which changes we need; how quickly we determine how to create the solutions; and how quickly we implement them. After strengthening our own IT competencies, we can change the model of interaction with contractors. We like the Time & Material model, when we set a task and attract qualified staff to solve it, and we also can flexibly manage these employees.
The more you have, the more you want. We want to implement more and more systems. For example, logistics specialists previously wanted to adopt just one system, and now there are 12 of them. And a specialist from the IT department needs to work this out very quickly. Therefore, IT has many new tasks that are overseen in two ways: one part is focused on existing functions and systems, and the other deals with new ones. Associated with this were the internal reorganization of the IT team and the division into several areas: development of information systems; development and operation of infrastructure, core services and communications; service and IT process management. Each function now has its own business partner, and each business division essentially has its own IT team. This is how the IT and Business Information function was created. Vladimir Savkin manages it. Vladimir has a lot of experience in the adoption, development, and support of ERP systems; he previously held top positions in large telecommunication companies such as Vimpelcom and Megafon.
As for insourcing at Sibur IT LLC, such an approach is a convenient way to manage. There are many legal entities in Sibur that can be established or disbanded for the sake of management convenience. If, for example, all Sibur IT LLC employees are transferred to Sibur LLC, this won't have a fundamental impact on anything.
In addition, we have the Digital Technologies function headed by Alexei Agapkin. Alexei knows the company's business well; he was head of Sibur-Neftekhim, and also headed the Center for Production Development. This function consists of four divisions. The first division — Advanced Analytics — deals with models. An experienced and attentive operator always manages the process more efficiently than a novice, simply because they have seen a lot more units in different operating modes: cold, warm, one or other raw material. They have significant empirical experience that will help to select the correct mode. Imagine that you were given the task of driving a car at 60 kilometers per hour and you cannot deviate by more than one kilometer per hour. But at the same time, the road turns, goes uphill, then descends, and the wind blows from the front, then from behind. However, if you drive at a speed of 55 or 65 kilometers per hour, then you will either get a poor-quality product, or you will simply be wasting raw materials or energy. But a properly trained model can take into account all this complexity and variation. As a result, it will be even more efficient than an experienced operator.
The second group is called Process Digitalization, and it consists of our internal developers. With the help of reengineering and application deployment, we are rebuilding and transforming a large number of processes so that our employees do not waste time on routine tasks. We already have a system for issuing electronic work permits, a solution for "mobile" equipment inspection, which appeared before the large-scale transformation by our IT, a sampling tool with registration in a mobile application, and so on.
The third division is called Industry 4.0. They work with new sensors and bracelets, test, and implement them to solve our problems.
An important feature of petrochemistry is that this industry deals with hazardous production facilities. Therefore, everything must be fire- and explosion-safe. There is a permissible concentration of hydrocarbons in the air, which can become explosive if a spark arises in a device. So, everything in these zones must be explosion-proof. But, for example, such sensors did not exist at all. Nowhere, not even in China. We are looking for partners and managing the development of such unique technologies.
The same group deals with machine vision. For example, there may be congestion on the conveyor for rubber briquettes. The camera will understand this immediately. Or, say, a pipe bursts or a flare becomes too bright. Or here's a real example: the rubber crumb on the conveyor is black, and it should be white. This means that there is some kind of external impurity. If you leave everything as it is, you will get a poor-quality 30-kilogram briquette, which will either lead to the customer having defective products, or, as is most often the case, it will be detected by employees who inspect the final briquette. They will have to cut out these black impurities. If there are a lot of them, then the whole briquette will be recycled. However, the camera can see these dark inclusions on the conveyor and initiate the launch of a rejection system.
In addition, we have a fourth small division that tests technologies that are not widely used in our industry. For example, there is a robot for testing the quality of polypropylene in the central factory laboratory. It has already successfully passed the testing phase, and we are waiting for a working prototype to be made for us. Is it cheaper than manual labor? We need to make calculations. There are, however, services that are definitely easier to buy. For example, drones patrol our product pipeline in Siberia (more than 1,000 kilometers) and detect leaks. We are testing many of these technologies for independent implementation; for example, there is a drone that can inspect the state of the flare (as a rule, this is the highest point of the plant) and other hard-to-reach places. In addition, there is a drone equipped with a thermal imager to help detect leaks or insulation problems. An infrared vision drone does an excellent job.
Of course, there is a dilemma: if there are any distant areas where you don't want to send a worker, then you can either put a sensor there or send a drone. Then we analyze and calculate which option is the cheapest.
At construction sites, we use a drone to inspect earthworks and soil movements. The system can report what happened at the construction site, as well as where and how much soil was moved. In addition, the system will compare this with the design drawings, which is very important. In particular, this software is often used in the USA for settlements with construction companies: moving one ton of soil costs this many dollars – here's your receipt.
Implementation of digital products, whether advisers or drones, is impossible without structured processing of corporate data.
Therefore, we have a third function called Corporate Data Management. It is led by Alexander Ayvaz, who is actually the Chief Data Officer (CDO). Over the past 10 years, he has worked with data storage systems and has been involved in the implementation of analytical approaches in various companies. We didn't have such a position before, but, starting digitalization, we realized that we couldn't do without it. How does it work? A manager makes decisions based on some data. The supporting system should provide the right data at the right speed, with the right quality and with the least effort. It should not be like this: several people enter data into spreadsheets at your request and comment on them, then another employee in the plant executive office collects data in a new spreadsheet and also comments on them, then he/she sends the data to the corporate center, where another employee collects the same spreadsheets and comments from several plants. As a result, it turns out that four hundred people participated in collecting information, and some of them made mistakes, so part of the data is incorrect and it is not clear how to interpret it now. How it should be: part of the data enters the system from the sensors, and another part is entered directly into the web form. The corporate data management specialists have already performed all the checks and normalized the results from inappropriate numerical ranges, after which processes with the maximum degree of automation are used to aggregate all the data and transfer them to the dashboards immediately.
We had an interesting case in a business division that produces polypropylene films. It has a proprietary and fairly rare process control system. And we couldn't find a way to download the data from there, so we had to ask the contractor for help and pay money for this work. Of course, when we have a data lake running, which is currently being built by the corporate data management unit, such processes will be significantly accelerated. The lake will be integrated with all major systems, and the speed and convenience of access to data will increase many times. And all dashboards will be filled automatically, rather than manually, as is the case now. This is an important part of our digital transformation.
The second important task of this function is to save Sibur employees from routine work. We want our employees to focus on intellectual labor.
How has the IT budget of these three functions changed since the end of 2017?
If I'm not mistaken, we have been growing for the last seven years thanks to an increase in the number of implemented systems. At one point, we decided to implement a document management system — an MES system — everywhere. The simplest pre-project survey showed that the organizational structure itself was not ready for this, and we had to change a lot to meet the new requirements. Then, other systems began to appear. And now, new systems need new features. Therefore, the growth of our total IT budget, as a rule, was two-digit (in percent), which was an expression of our aim to actively adopt new systems.
I should emphasise an important point: a few years ago, the IT function employed about 800 people. And then their number increased to 1,200. However, at the same time unit costs decreased. This happens when companies recruit people instead of getting a service on the market. It is also important to interpret the budget: when you have a payroll fund, and when you bear the costs of third-party services.
I would essentially divide the IT budget into two parts: run (maintaining existing systems) and change (launching new systems). The run budget is growing slowly, especially since it is constantly being optimized. We analyze how many licenses we need, how to buy traffic cheaper, what we can do on our own, and what is better to buy as a service. Such optimizations in this part of the budget can prevent or slow down growth. But the second part of the budget is undoubtedly increasing. However, the change budget is not distributed as an IT budget, but as a budget of what we call the "organizational project." We have a special project management committee that reviews the feasibility and potential business benefits of implementing a particular system. When it makes a favorable decision, part of this money is allocated through IT.
Is digital transformation more about insourcing than outsourcing?
I am convinced that when you buy a service, you still need to understand it well. Otherwise, you risk not getting exactly what you need, and you will also overpay. If you do not have people who understand well how to properly develop, build models with big data, and so on, then you will not get the best result.
How many people are employed by the information technology divisions at Sibur?
About 1,500 in all three functions.
You recently said that you are "starting to solve a class of problems that few have solved in Russian industry yet." Can you provide some examples of such unique projects.
For example, explosion-proof sensors, which I have already mentioned. They just don't exist yet. There are also robots, but, as it turned out, setting them up takes too much time and money. That was a surprise for me: Many companies began to deal with this problem before us, but for some reason it did not occur to anyone to develop these sensors themselves. More precisely, it hadn't occurred to anyone to develop cheap options, because there are expensive ones, but their cost is three-digit, and we need tens of thousands of such sensors. Therefore, we need an alternative product that will work using a simple battery. We are creating it ourselves.
Another class includes tasks that are not unique as such, but which are unique to our company. The code that we write will naturally be one of a kind. Our mobile patrol inspector will take into account our unique needs and will not resemble any other mobile patrol inspector, despite the fact that there are a lot of them. And it will do exactly what we need. Digitalization cannot be likened to installing WhatsApp — that is, you cannot take a fully packaged product, set it up, and immediately get the desired result. The number of customization tasks is always vastly greater. Or there is an tool-tip for railway logistics. It helps us to optimize repairs and analyze rolling stock needs; it also suggests dual operations and advises us how to consolidate shipments. It predicts cost-effective situations and simplifies logistics operations.
In addition, there is a very ambitious project related to off-site control of remote plants. Operators are usually located in an explosion-proof bunker a few hundred meters from the production plant, and control commands are transmitted over an optical fiber cable. As for oil platforms, there are examples of control from a distance of several hundred kilometers, which means that companies do not need to send shift workers to the oil platforms, but manage them from the coast. In the industrial gas production sector, there are dozens of plants operated from the central control room, which may be located in another region.
Such an approach is not currently used by large polymer enterprises, but the experience gained in other industries and the available tools allow it. We are considering such innovative solutions for our new production facilities.
What business results have you already achieved with the help of new technologies?
Our cost reduction has two components: either money or man-hours. For example, an employee used to spend an hour and a half to get a work permit, and now it only takes ten minutes: we freed employees from the routine tasks of manually filling out documents. If you count all such benefits, then the cost reduction in the current portfolio is about 10–15%.
The second component of cost reduction has a monetary value. For example, advanced technologies allow us to produce more butadiene and earn a profit from this. And the digital twin of our railway logistics system helps save money on wagon repairs or on freight charges. Based on the experience of implementing similar projects in other companies in our industry, these activities can provide savings of up to 10% EBITDA. And with each step forward, we see more and more new opportunities.
You mentioned the popular technology of digital twins. Is this a priority for you?
Everyone that we ask has a different take on this word. So many people, so many minds. I believe that the digital twin is a mathematical model — i.e. a function. It's just a kind of software that describes the behavior of a parameter depending on some other parameter. And it can be very simple or very complex, depending on the components "under the hood."
Take for example, the big data model: in a nutshell, it allows you to use the analysis of data collected over several years to predict a particular situation. For example, I can equip a car with sensors for wind speed, slope and turning radius and ask you to ride different roads on this car for a year, after which I will say: "I have no idea why, but I see that with such an impact on the pedal, with such a wind from such an angle and with that speed, your car is driving at such a speed on a slightly slippery road." And it has been in this state five times. Now, I know for sure that the car will drive at such a speed under such conditions. Therefore, if you need the car to travel at a speed of 60 kilometers per hour, you need to press the gas pedal with a certain force and turn the steering wheel by a given number of degrees. But if you need to drive a car in a mode in which you have never driven, this model will not help you.
But a mathematician does not need historical data. Knowing air resistance, friction coefficient, and other parameters, he/she will create a formula, load it into a computer, and get an answer. This is a different model, and it is being conducted by other specialists. But they are both digital twins. And each facility can have a huge number.
In one of our cases, we had to calculate how many tons of certain products we would get if we burn the catalyst at a certain speed. You can burn the catalyst quickly, face excess overloads and expenses, but produce a lot of products. Similarly, you can burn the catalyst slowly, but when it is time to change it, then it may not be necessary. Therefore, you need to ensure balance. And this is what one class of models does in Sibur. In addition, another class of models for the same process unit can predict how the furnace is coked. And you will be able to manage your furnace more effectively.
Each of our cases with a business effect needs its own model.
A terminal which scanned my passport and took a photo issued me a biometric pass to the headquarters. Is this technology used throughout Sibur?
It is not currently used everywhere, but soon it will be. To do this, we need to continue changing the pass entry system and creating the necessary conditions.
You have been using virtual reality in staff training for quite a long time. Where else do you use or plan to use VR?
This practice remains somewhat localized, because at the moment it is very time consuming. But we are implementing several more use cases. For example, ZapSib uses it for operators who perform a rare and dangerous operation with a highly flammable substance.
It seems to me that these technologies will be more widely used when the entire process of transferring information from the designer and vendor to the operating personnel is configured. We are now implementing such a pilot project in our subsidiary NIPIGAZ. You need to take care of this at the very first stage of the life cycle of an enterprise — during the design process. And you immediately explain to the designer that he/she will have to provide such drawings in a given format with a given degree of detail. The same applies to those who will sell us equipment: they must give us all the information, drawings, repair strategies, data sheets for facilities, etc. This is necessary so that we do not spend time filling in the equipment database for describing and creating hierarchies for the databases of repair standards and maintenance. And for us, it's important to have 3D models from the very get-go. That is, we need to get the data into our storage directly from contractors, after which we could create many 3D models using special software. This is also important for training purposes. We shouldn't use a scanner and spend a lot of man-hours on it; rather, it should be a smoothly running conveyor. And then it can become mainstream.
Could you tell us about your plans for the use of wearable devices?
Basically, these are industrial mobile phones for two types of employees — equipment operators in the field and repairmen. Employees of the first type use them to record the status and malfunctions of equipment and then instantly transmit information if something is wrong. It also helps to reduce the number of log entries because the information can get there immediately. Employees of the second type register the beginning and end of each operation and may also receive drawings. The list allows them to select a tool and so on. This is also an excellent source of useful information.
We have AR glasses, which we are testing now in Tobolsk and Nizhnevartovsk; and we are also testing bracelets and T-shirts.
What are your plans for developing your own infrastructure? Do the company's existing data centers meet your current needs?
We are gradually changing our infrastructure. When our big data analysts need a server, we give it to them. In addition, we have already bought two server racks for a data lake.
What do you think about import substitution? How actively does your company use domestically produced IT solutions?
I can't say anything new. There is a known problem with databases that has not been resolved yet. No one has created Russian solutions of the quality and scale we need. However, on the other hand, we are developing some programs ourselves. And another important issue is risk management. This is especially relevant because we use cloud servers that are located abroad. When we discuss the possibility of using cloud servers, we always ask ourselves the following question: "What if…"
Does the country of origin matter when choosing a technology for implementation?
We will be happy to consider any Russian solutions when they start to appear. We are even discussing preliminary options with some integrators.
Innovations require the appropriate human resources. How do you find them and which professionals do you have to train yourself?
As I already said, many digitalizers view us as a strange company that bucks the trend. We can't simply advertise and wait for specialists to come to us – such a strategy doesn't work well. In fact, new divisions were established due to the recruitment strategy, which was developed jointly with the HR Department and was not used very widely before. The strategy included hackathons, workshops, and professional events. And this approach really works. We still have to overpay for being unusual, but it seems to me that over the course of the year we have become much more mainstream. And, of course, we strive for greater brand awareness among IT professionals, which will help us with recruiting. We hired 80% of the staff we needed in 2018.
The next question is a little philosophical. Many people are now so deeply involved in digital transformation that they are talking about actually transforming into an IT company. Others have a different point of view and believe that they should focus on their core business. Is Sibur going to become an IT company?
In a broad sense, of course, IT assumes an increasingly important role, but petrochemistry is an industry that is associated with several very specific things. It has always been like way and always will be.
First, we need to get cheap raw materials and make the molecule in the most efficient way, so that, in terms of logistics, this molecule can get to the market where it is needed. This is the foundation on which the Middle Eastern petrochemical industry was built. Cheap raw materials have stimulated the development of petrochemicals in China; a huge industry has emerged there that uses polymers and manufactures finished products for various consumers. The nearly extinct petrochemical industry was revived in the United States, because they found a way to get cheap ethane thanks to the shale revolution. And this is an exceptionally important topic. IT would never have revived the US petrochemical industry. The fact is that the industry found cheap natural gas from which they extract ethane.
The second important component is the ability to produce molecules with new properties that were previously unknown. Digitalization helps chemistry, but the foundation is still the knowledge that chemists have about how to make very lightweight and durable plastics that will replace the standard 100 kilos of metal in the frame of a car, which will reduce fuel consumption.
In addition, IT in petrochemistry is not equivalent to IT in banking, where you can save people from having to visit bank offices, assess creditworthiness remotely, and also offer the right products at the right time and place.
However, companies that do not engage in digitalization will greatly slow down their development, because they will not be able to leverage many of the existing opportunities. Sooner or later they will lose competitiveness and face an outflow of customers. Therefore, I believe that digitalization will be present in all industries, and it will be mainstream for them.