LKS Next – Smart press optimisation system
LKS Next
Sector: Industry
Business Case
Pressing machine settings in metal stamping and drawing processes are adjusted through trial and error. This method is slow, costly and sometimes inaccurate. For this reason, and in order to prevent failures such as necks, wrinkles or fractures from occurring more quickly and efficiently, the use of Machine Learning is proposed for this press parameterisation process.
Objectives
Develop a system capable of reducing the defective material resulting from the stamping/embossing process of metal parts through the use of artificial intelligence models, capable of determining the optimal parameters for the stamping/embossing machine.
Use case
Using historical data, a classification algorithm is trained that can calculate the probability of breakage of the parts to be manufactured. Subsequently, the optimal parameters for the fabrication of the parts will be determined through a genetic algorithm that minimises the probability of breakage of the sorting algorithm.
Infrastructure
On Premise and in the Cloud
Technology
RPA
Data
Data on the manufacturing processes of the metal parts: machine used, parameters of the pressing machine, part being manufactured, place where breaks occur, physical-chemical characteristics of the materials used, etc.
Resources
Multidisciplinary team composed of data scientists and programmers.
Difficulties and learning
The difficulties faced consisted in the use of a database without a suitable structure for the purposes of this project, and in the lack of knowledge of certain services of the cloud platform used (which required a previous learning phase).
KPIs (business impact and metrics of the model)
The classification model is able to determine the percentage of error that the manufacturing process will have, so workers will know what preventive measures to take, thus increasing production quality. In addition, the genetic algorithm determines the parameters that minimise the generated defective items, thus saving material costs and manufacturing time.
Funding
This project was funded by Sodercan, Cantabria’s project funding.
Collaborators, Partners
BSH
