Success story I

  • Home
  • /
  • Success story I

Statement

“Use of Legato to apply Machine Learning techniques in the construction of a predictive model for the manufacture of MDF boards”.

This project exemplifies the value of Legato in an industrial environment, both for the improvement of productive efficiency and the final quality of the product.

For what reason and what purpose?

In industrial production plants, the control of the multitude of variables that intervene in the processes and determine the quality of the final products is critical. The control of these variables is not a simple task and requires continuous adjustments by plant operators. In this case, variations in factors such as belt speed, temperature or humidity cause frequent deviations from the set point that marks the optimal behaviour of the process. Control and prediction, therefore, were critical.

The objective of the project was to model the behaviour of a limited phase of the manufacturing process of medium density fibreboard, so that its final quality was adjusted to previously defined quality specifications and increased production efficiency. Both objectives were achieved.

Employment of Legato.

In this project Legato was used for data analysis and to develop the prediction of the behaviour of the manufacturing line. In the first phase we work with historical data (elaboration of the theoretical model), in the second phase real data is exploited and predictions made at the pace required by the manufacturing line.

Initially Legato worked in depth with a set of about 60,000 records corresponding to some 70 different variables (weights, pressures, speeds, levels, heights …) collected over a week. With this historical data, Legato determined which were the variables that, in the manufacturing phase object of study, were more determinant in the quality of the final product. For this it was necessary to apply a set of Big Data technologies that arrange, classify, clean and qualify the data. Subsequently, the transfer function that describes, over time, the behaviour of the process was calculated.

Once we had obtained this valuable new information, we already knew how things had been done, what the result had been and what factors and to what extent they had intervened in it. We could then begin to determine the criteria for improvement.

Thus, Legato developed a scorecard that offers all this information in real time. Then, after an intense phase of testing and selecting the most efficient combination of algorithms, applying Machine Learning technology, it introduced in this scorecard the descriptive graphics of its prediction, indicating to what extent a certain variable should be adjusted, identified as critical, to obtain a result very close to 100% reliability with respect to what had been previously defined.

Conclusions.

Once this work had been done, through our analysis and prediction engine, the model has started to propose recommendations that allow decisions to be made to improve the productive efficiency and the final quality of the products. The implementation of this solution ends with the automation of the actions derived from the recommendations proposed by Legato. In other words, Legato achieves that the line itself can self-regulate, obtaining better results in quality and costs.