Resumen
We have designed and investigated a system of automated control for a carbon baking furnace that makes it possible to minimize the probability of defects. Based on the artificial neural networks, it differs from actual systems by the possibility to select the technique to control the process of baking in accordance with the starting conditions and the purpose of the process. Such purposes in the process for baking carbon articles include the obtaining of products with a minimum number of defective structures or a safe reduction in the length of the technological process to save energy.To minimize the number of defective structures in finished products, it is proposed to predict the probability of a defect in a product and, to reduce the duration of the technological process, to apply entropy as an indicator of product readiness. Solving the set tasks is based on the use of artificial neural networks with their capability to generalize data, namely different values of the probability of a defect in the various temperature fields in a baking furnace under operational modes.The issue of the limited amount of data required to train an artificial neural network is resolved by applying a special structure of artificial neural networks, an autoencoder.The designed control system has several advantages over the systems already in use. For example, it makes it possible to select a step of the descent that determines the accuracy of the optimal descent trajectory and, therefore, the accuracy of control in general. Selecting a descent criterion makes the designed system flexible when used under different conditions and for different control tasks. In case of an emergency interruption of the technological process of baking, a given system makes it possible to plan its further progress at any time, thereby providing for an effective continuation of the baking process and thus avoiding unjustified expenditures. The control algorithm used in the system makes it possible to predict the duration of baking and the value of fuel consumption and thus predict the economic efficiency of the technological process.To assess the efficiency of the proposed system of automated control over the baking process, we have performed an experimental study of the designed system operation compared to the actual control system based on a PID-controller.The study results have demonstrated that the use of the new control system makes it possible to reduce fuel consumption by 3?4 m3/h. In addition, there is a decrease in the growth of the temperature of blanks during treatment, which positively affects the quality of the finished products