Business Case
Cerámica San Lorenzo (CSL)
Case Study AI Bruna - Powered by Altumlab
Cerámica San Lorenzo (CSL) is one of the main ceramic tile producers in Peru. They supply part of the Peruvian, Chilean and Brazilian markets. In one of its plants there are production lines, where it is possible to configure the production of one or more product sizes according to the need.
To produce ceramics, a mixture of clay and other elements are subjected to a moisture treatment giving it cohesive properties, which together with a procedure where it is subjected to a high pressure load forms the required size.
The raw ceramics, with their characteristic reddish color, are then routed through a belt where they will pass through the different stations where paint bases, colors, and finishes will be applied until they reach the packaging area to be stored before being shipped.
The different products are classified by product family, finish and their respective size or format. The production lines must be configured each time a product is changed from one to another, however it is possible to reduce the configuration time when two products that share the same size are produced consecutively, for example, in order to use that time in production.
The logic of coupling "similar" products must consider multiple production and commercial restrictions, such as minimum volumes to be produced each time a SKU is produced, channel and customer satisfaction considering that there are delivery deadlines, international shipment deadlines and demand forecasting by product, by channel and by market.
In the plant there is a planning team, which builds and adjusts what or which product should be produced in each of the lines, this being an exhaustive and highly complex task, since there are daily changes in demand that impact on the previously established production.
To solve this, CSL's Head of Planning initiated the development of a strategy that would allow the use of Artificial Intelligence to solve this problem. Since there was no solution available in the market, the AltumLab Research and Development Center proposed a possible solution, where considering the commercial demand and the different operational restrictions, production scenarios are projected for each line.
To start with, the project covered one plant, considering uniform and multiformat lines. With this we started to develop a neural network to optimize the use of lines considering all operational constraints, days and hours of operation, production costs per m2 per line and the minimum satisfaction per channel in the demand forecasting.
The planning of an entire plant can be executed constantly on demand, where data is consumed directly from different sources and systems such as ERP, WMS, DB and similar, saving significant time in the ordering and collection of information.
The preliminary results of the solution allow a 4% increase in production with the same facilities, but with a dynamic planning system, where considering each sale made, it is possible to re-plan in a matter of minutes the destination of each line, generating additional value to the available assets.