R+D CSIC

Scientists develop an affordable artificial intelligence tool to improve the production of SMEs

The CSIC and the Polytechnic University of Madrid have developed a software based on machine learning techniques to generate useful information and intelligent recommendations to optimise production processes. This affordable software has been created within the framework of a European project to give SMEs access to this type of tools.

"Companies don't always know which factors have the greatest influence on results," explains Rodolfo Haber, a researcher at the Centro de Automática y Robótica, a joint centre of the CSIC and the Universidad Politécnica de Madrid (UPM). "The companies set certain parameters on the production process and think everything it's going well. But if they would analyse their data properly, they might find that by changing the parameters, results would improve.”

"One of the most greatest value of companies is their data", Rodolfo Haber says, “and data analysis can bring many improvements for the company.” However, such analysis can be costly and complex for small and medium-sized enterprises.

In this context the European KITT4SME project was launched, aimed at providing SMEs with affordable artificial intelligence tools to optimise their production.

Within the framework of this project, the research team led by Haber has developed a programme based on machine learning techniques to analyse all the parameters and variables of a manufacturing process.

The software compares all the data with economic and production indicators, develops various optimized models, and selects the most suitable one for the company. It also suggests changes to parameters and variables that can lead to improvements while quantifying the resulting benefits.

"One of the most greatest value of companies is their data and data analysis can bring many improvements for the company."

The programme can be applied to any production process regardless of the level of technological maturity. It utilizes production process data, machine parameters, and performance-related variables (such as production rates, rejection rates, quality, and time), making the most of historical data.

With all this data, it performs automatic modelling using artificial intelligence, optimises the parameters of the models and selects the best of them. It then uses an evolutionary algorithm to enhance the selected model and determine the most relevant variables and parameters. The higher the degree of technological maturity, the easier it will be to implement the tool, researchers say, although the company will have to do some initial work to introduce and condition the data.

The researchers are looking for more companies willing to collaborate to further validate the tool before proceeding with commercialization.

Succesfully tested with a pilot line an a company

"Current tools," adds Haber, "are very expensive and only accessible to large companies. Current IA commercial tools almost always perform such an assessment in different stages and the results are neither interpretable nor directly usable by companies, which creates a high dependency on experts in the field and very long processes, with costs that are unaffordable for many small companies.

The tool, still in the validation phase, has been tested with data provided by a pilot line and a company, with successful results. The researchers are still looking for companies willing to collaborate to bring the solution to a level 7-9 on the technology maturity scale, in order to further validate the tool before taking it to commercialisation.

Contact:

Marisa Carrascoso Arranz
Vicepresidencia Adjunta de
Transferencia del Conocimiento - CSIC
Tel.: +34 915681533
This email address is being protected from spambots. You need JavaScript enabled to view it.

• Prev
• Next