How The Grain solves production scheduling problems in minutes with Hexaly

Case study Manufacturing Scheduling

Léa Optimization Scientist October 21, 2024

The Grain is a software company expert in solving production scheduling problems for various industries. They build solutions to provide businesses with real-time insights and actionable recommendations, helping them reach their operational goals. The Grain switched to Hexaly in 2 days and now delivers optimal scheduling in minutes to their clients.

Various production scheduling problems

The Grain faces complex and challenging production scheduling problems with many different types of business rules and goals. They have clients in various industries, such as the process industry, discrete manufacturing, and assembly. Every industry is different, and each company’s problem has its specific business rules and goals, ranging from maximizing production output while dealing with bottlenecks to complex changeover matrices, scarce capacity, resource balancing, and people skills.

For example, in the process industry, the goods must first be processed and then packaged, which creates complexity. Another striking example is assembly, even for small companies. They typically create hundreds of different products from thousands of components and raw materials. In this case, the most critical challenge is keeping track of inventory. While these problems are known to be very hard, The Grain manages to solve them all efficiently using Hexaly. When computing a plan manually used to take days, it now only takes minutes.

Quick, optimal scheduling using Hexaly

To tackle all these problems, The Grain designed a tool based on a set of template solutions for specific industrial production processes. Since each business is unique, the standard solutions they offer their clients must be configurable to fit their needs. To do that, they built a semantic layer to write the configuration, using business terms the clients understand, on top of the mathematical formulation.

Thanks to Hexaly’s innovative nonlinear and set formalism, the mathematical formulation is easy to write. Moreover, Hexaly’s use of intermediate expressions allows the writing of compact models. For most problems, the only decision variables represent the starting times of production orders. From these decisions, one can define every other quantity necessary to express the constraints and objectives of the models. As a result, Hexaly models are much more straightforward and transparent than classical Mixed-Integer Linear Programming (MILP) formulations.

The size of the instances they solve is very large. Indeed, an average factory has around 15 packaging lines, 30 mixing tanks, and several thousand products and inventories to manage. The planning horizon typically ranges from three days on a 10-minute basis to a year and a half on a weekly basis. This instance size results in models with 50,000 to 100,000 decision variables, which Hexaly solves in minutes or even seconds. The solver’s great performance on these problems allows planners to use the tool to build schedules for the next day or week and test different scenarios. For example, if a company plans to increase its production capacity in the following years, it can quickly run several experiments to determine which equipment or parts of the factory it should enlarge.

Why Hexaly?

The Grain’s team was convinced by Hexaly’s great performance in solving their planning problems. It not only provides near-optimal solutions but also does so very quickly. They can now solve even their largest instances in minutes. Before Hexaly, they relied on other commercial or open-source solvers, but their performance did not meet their expectations.

Besides performance, the team was also very impressed with Hexaly’s ease of use. Despite being a new modeling formalism, they could switch to Hexaly in a few days. The Grain team greatly appreciated the dedicated and responsive support provided by Hexaly’s optimization scientists, who helped reformulate their models to achieve the best possible results.

We were very impressed with Hexaly speed in making things happen! Actually, we had a solution running on another solver for a client, and we switched to Hexaly in less than two days, which is amazing. And it worked! We also encountered a few things that didn’t work initially in the model’s building. With the support of the Hexaly team, we could figure out why things were slow, and we reformulated the problem appropriately. The model is much smaller with the new formulation, and now it’s flying! Hexaly really helped us get orders of magnitude faster and better solutions. We fully recommend it! 

Steven Raekelboom Technology & Innovation Director