Developing an analytical system for a tire manufacturer
Pirelli & C is one of the world's largest manufacturers of premium car tires. The company owns 24 factories in 13 countries, including two factories in Russia.
In 2020, the Voronezh Tire Plant, owned by Pirelli & C, decided to develop a manufacturing analytics and IIoT system for analyzing the enterprise's production processes. But ready-made solutions "out of the box" could not offer the necessary functionality. So the company turned to Evrone with a request to develop a unique data analytics system for manufacturing production, which would allow the data to be analyzed and visualized and let them see bottlenecks.
The challenge: collect data from hundreds of machines in one place
The plant has hundreds of machines that perform millions of operations every day. At the time, all the raw data on them were sent to one large common system. Each department unloaded and processed it manually, which took a very long time. While they could unload data from the system via SQL, this had to be done by specially hired specialists.
Then each department prepared a report according to its profile: planning, quality assessment, etc. As a result, top management collected several different reports and, based on the data, received a certain general picture of production.
But thousands of such reports were collected: for a shift, a day, a month, etc. To create the annual report, employees in each department had to manually process huge amounts of data. This took up most of their work time.
The plant needed to develop an industrial analytics platform that would increase the efficiency of operational processes, take over the primary analysis and processing of data, and issue it in an easy-to-understand form, using graphs, tables, and diagrams.
Basic system requirements:
- Ability to work with different data sources
- Integration with enterprise information systems
- Reporting platform
- Flexible output of results, in the form of graphs and tables
- Authentication and authorization of system users
- Deployment within the international corporate standards of Pirelli
The solution: analytics and cluster structure
We had to come up with a unique industrial data management platform, so we did a lot of preparatory work before beginning the manufacturing monitoring system development. In the first months, two Evrone business analysts worked on the project, studying architecture and use cases of existing databases and user requests for a future product.
The technical team also visited the enterprise and got acquainted with the technological cycle and equipment, in order to understand how production works and the existing internal bases and systems.
Currently, one analyst is working on the project. He is engaged in clarifying the needs of the client, collecting feedback on ready-made solutions, and formulating tasks for developers. All communication with the client's technical team, most of which is located in Italy, goes through him.
Big data in industry
We needed to develop an application for the collection and primary processing of big data in manufacturing. The company has an industrial Internet of Things (IIoT) system. In addition to data on the direct operations of the equipment, operating times and types, the number and duration of cycles, and failures, the machines also send data about various rubber compound formulations and tire models to the databases. This information flows into local databases and then accumulates in the general Data Lake Pirelli. Management and access to data occur through the internal system of the corporation.
We taught the application to quickly find the necessary data in the storage, filter it according to the specified criteria, and then build tables, graphs, and charts based on it. But the application does not just collect archived data; it is also a real-time production monitoring system, allowing users to monitor current processes and report failures and violations. In addition, users can correct the results, in which case the edited version is saved separately.
The main platform for launching and managing running applications is Kubernetes. We chose this platform for several reasons. One of the main ones is the ability to flexibly and quickly scale environment resources when the load on services increases, especially when the number of analytical modules in the system increases. Built-in isolation mechanisms allowed us to work with data safely and accurately. Colleagues from Italy have already transferred some of their services to this platform, and we quickly optimized our deployment processes to work in the global Kubernetes environment, thanks to unified approaches and tools supported by the global DevOps community.
The distribution of the application is implemented through Docker, which allows us to quickly deliver changes to the production environment using the optimal development languages for specific tasks, as well as to flexibly switch between releases. The entire process of building and delivering releases is automated using CI/CD tools and services, which eliminates the influence of the human factor in these processes.
The application is written in Python using Django, and React is used on the frontend. To work with databases, a layer in the form of a Query Engine is used, which allows you to flexibly manage access to tables, as well as optimize and parallelize queries when working with big data. Caching of frequently accessed data is implemented using Amazon ElastiCache.
The request for the interface design of the BI tool came from our development team, not from the client, as is usually the case. The client was not satisfied with the initial prototype using ready-made solutions — there are too many different forms of data in the project, and it was extremely difficult to find the information of interest in the final graphs using existing BI tools in the manufacturing industry.
It was necessary to take into account some conventions that are already accepted in the company. For example, color coding. Conditional colors are used to indicate different groups of data. There are also preferences for chart types — most often, Pareto and waterfall charts are used.
We have written scripts for different user groups to make the system more convenient for different specialists to work with. You can choose the period for displaying statistics, the number of machines, production cycles, and so on. The most useful buttons have quick access, so that interface elements do not interfere with the perception of basic information.
Result: MVP release by modules
The implementation of this analytics system is divided into modules responsible for different aspects of production, which we are implementing in priority order for the customer. There are 17 parts in the project, and three of them have already been released into production.
The first module is responsible for monitoring the Phase In processes — the transition of equipment to new tire size and the production of that size. In this case, there are two types of Phase In — BU and CU, for which the sets of control points are different. To analyze these processes and optimize the performance of production, we taught the application to collect control points for all stages of production (assembly, vulcanization, and finalization) and compare them with normal indicators, so that any deviations are immediately noticeable.
Next, we implemented an OEE system that allows us to evaluate the performance of the equipment. It operates with data on the operation of machines directly, without reference to models and recipes. Data can be filtered by dates, shifts, types of machines, specific machines, or you can reset all filters and see a general summary table.
The third module allows you to track equipment cycles and the performance of operators. The system generates statistics on the time of work cycles for machines, rubber models, and operators.
We are still at the beginning of this journey, and we plan to add many other useful functions to the analytical system — for example, tracking product quality and maintenance optimization. As a result, the Voronezh Tire Plant will have its own automated monitoring and primary analytics system, which will help provide process efficiency improvements and save money, time, and resources.
If you are also looking for custom manufacturing analytics solutions or production intelligence apps for big data and industrial processes, send us a message, our experts will help you to develop a custom product based on your requirements.