Executive summary

Context

Atlas Copco is a global industrial company and world leader in air compressors manufacturing. Production planning on the assembly lines is crucial for the company’s activities, allowing the transmission of accurate and reliable information to employees and clients.

Goal of the project

Optimizing the production planning of Atlas Copco’s Portable Air manufacturing sites, where complex products are assembled to fit its customers’ requirements. The project aimed to make production planning faster, more efficient and accurate, while reducing human time spent and improving the number of orders delivered on time.

Solution

An AI solution that optimizes production planning at Atlas Copco. Superlinear developed a software that allows both Production Planners as well as Sales Engineers to quickly generate accurate and optimal production plannings.

Case Study

The challenge

Atlas Copco is a global industrial company founded in 1873, active in 180 countries and employing more than 39.000 people around the world. Atlas Copco is a world-leading manufacturer of air compressors, vacuum solutions, generators, pumps, power tools and assembly systems. The Portable Air division designs, manufactures and markets mobile compressors, including high-pressure boosters, and light demolition handheld tools for industries all over the world.

Innovation is in Atlas Copco’s DNA. The company conducts ground-breaking research in Wilrijk, Belgium, where most of the compressors sold throughout the world are designed and produced. Atlas Copco also focuses on sustainability, with major developments in smart factories and compressed air as a service.

Production planning is crucial in Atlas Copco’s activities. It allocates employees, materials and machinery while taking multiple, complex variables into account like items availability and production capacity. It is essential that the production planning be accurate, flexible and always up to date with the newest information in order to make on-site production as frictionless and efficient as possible. At the same time, customers need to receive accurate information about the status of their orders.

Production planning, therefore, requires a lot of time, resources and human effort. Atlas Copco aimed to optimize with AI the production planning of their air compressors’ balanced assembly line.

For this project, Superlinear worked with Atlas Copco’s Power Technique business area, where industrial ideas turn into leading-edge technology in air, power, and flow solutions.

The briefing

For its production planning, Atlas Copco was using a commercial solver, which has a steep licensing cost. These types of commercial solvers can quickly become highly expensive if they are to be implemented on different sites, divisions and countries, especially for global companies such as Atlas Copco.

Atlas Copco was looking for an AI solution that would optimize the production planning of the Portable Air manufacturing sites, where complex products are assembled to fit customers’ requirements. The optimization was to take place on multiple levels:

• Making production planning faster, more efficient and accurate.

• Reducing the human time spent and improve at the same time the number of orders that are delivered on the requested delivery date.

• Getting reliable and clear information to the client to deliver on promises, and allowing Atlas Copco representatives to better allocate their time. The goal was to give sales real-time insight into the capabilities of the production facility.

Furthermore, the solution needed to be:

• Flexible, as there are a lot of moving parts and complexity to production planning.

• Scalable: easily integrated into other sites and divisions of Atlas Copco.

• Implemented with 100% Open source software, so that the solver could be integrated into Atlas Copco’s own systems while allowing for easy continuous improvements.

• Developed fast and ready to be rolled out as soon as possible.

Last but not least, the AI solution needed to check all the boxes and integrate the constraints of the assembly line:

• Type and number of necessary items, availability of said items, planning, inventory capacity, employees, and machinery needing to be involved, etc.

• Production planning is continuously changing due to some orders taking more time than foreseen or new orders coming in. The solution needed to be customizable and allow room for fine-tuning of the parameters.

“Superlinear quickly understood our specific challenge and offered us a solution which helped us to optimize our production planning. What distinguished Superlinear is that they’re pragmatic and efficient, but not afraid to get technical when they need to. Looking back, Superlinear offered us a flexible and customizable API which slotted seamlessly in our existing processes – ticking every box for us.”

Bart Vercammen - Team Leader Manufacturing Systems & Processes at Atlas Copco

The work

Superlinear, in collaboration with professor Seppe vanden Broucke (KU Leuven), was able to develop the first sprint of the AI solution after just 2 weeks. That first version included a fully-fledged mathematical and AI model.

Superlinear and Atlas Copco set their sights on COIN-OR CBC, an open-source solver for the project. COIN-OR CBC is very versatile, but needs to be meticulously tuned to the specific task that needs to be solved.

After 2 months of work and 3 iterations, Superlinear’s AI optimizer was ready to be fully rolled out. The solver offered a 100-fold speedup in 3 versions by iterating on the optimization problem definition, the solver parameters, and the way in which the solution was compiled.

The solution immediately showed strong results: Superlinear’s AI-powered optimizer spends 91.6% less time than a human planner, improves the number of orders delivered on time by 9.0%, and reduces the total number of days orders are late versus the first promised delivery date by 17.2%.

The software is flexible and can take all the different parameters and variables of production planning into account. It is also future-proof, easily extended to solve new applications for Atlas Copco. The solver speed is competitive with commercial solvers. The majority of problems can be solved in under one minute, making life easier for Atlas Copco.

Superlinear’s optimizer has successfully been integrated into a planning system developed within Atlas Copco. It offers real-time integration with the sales system: the sales representative can now give a lead time immediately, even while still in meeting with the customer.

It is used by the production planners in the product company who are each responsible for multiple assembly lines where they have at each moment in time several hundreds of order lines in the planning. The solver helps them to optimally promised delivery dates for new orders and to reevaluate whether the current planning optimally satisfies customer expectations and commitments.

Superlinear’s solution is currently implemented at 2 Atlas Copco production sites in Belgium and one in the USA. It can be modified and updated inside Atlas Copco’s own system.

“Superlinear’s solver takes into account all aspects and constraints of the production planning challenge in a fast and effective way while offering real-time accurate information to the sales channels. It enables us to communicate faster and better with our customers, formulating commitments we can keep. At the same time, the solver increases the effectiveness of our customer contacts. The extra freed-up time can be used to focus on the planning of new projects or further optimize our processes.”

Bart Vercammen - Team Leader Manufacturing Systems & Processes at Atlas Copco

The future

Superlinear’s solution is planned to be rolled out in other countries and divisions at Atlas Copco.

“As the solution is open source and fully parameterized, it makes it possible to easily roll it out to our other production facilities in Belgium, US, China, Brazil and India where Atlas Copco product-specific focus might result in different requirements. Superlinear was very flexible in the support they gave us. They even took the time to show us how to tailor the solution ourselves for further applications within Atlas Copco, that was much appreciated. All in all, we were very satisfied with the enjoyable and efficient collaboration with Superlinear. We look forward to working together with them on other innovative AI solutions.”

Bart Vercammen - Team Leader Manufacturing Systems & Processes at Atlas Copco

“We're very grateful to Atlas Copco for having given us the opportunity to improve their production planning, and we're thrilled to see that the solution provides a significant impact on the number of orders delivered on time while at the same time making the planners' jobs easier. We're excited to see in what other areas AI can deliver impact for Atlas Copco, and already have some interesting low-risk but high-value ideas in the pipeline!”

Laurent Sorber - Co-founder and CTO at Superlinear