Product performance: A prediction model for compressive strength of composed cements

Abstract

As a result of a production process, it is important that the product meets the quality requirements defined by standards and specific customer needs. In order to verify the fulfillment of the requirements, the application of techniques to assist the evaluation of parameters throughout the production process and identification of possible deviations is important. Therefore, the available data need to be transformed into useful information and decision value, so that conclusions can be reached in a timely manner for possible maintenance and corrections. In the cement production flow, and in the consequent quality evaluation, it takes time between production and the availability of variable values. For this reason, the present study was developed aiming to predict, through the application of the Artificial Neural Networks (ANNs), the result of compressive strength at 28 days for the produced cements regarding parameters from the productive flow. ANNs have the ability to learn by mapping input data and their relationships with the output, as well as to synthesize and generate appropriate responses to a set of new inputs. The results obtained in this study led to a predictive model with a conservative profile and a strong correlation with the observed data, reinforcing the ability of this methodology to predictive problems, regarding the limitation of a small number of samples available for the training and the validation steps.