The quality of asphalt concrete mixture at the output of an asphalt concrete plant is unstable due to disturbances that we cannot control or control with significant delay. Disturbances may include factors such as inaccuracies in the existing relationships between the properties of asphalt concrete mixture components and the parameters of the technological process with the quality of the finished product. Disturbances can also be attributed to our lack of knowledge about the relationships between individual indicators and the quality of the mixture. Forecasting these disturbances to determine the actual quality at the output becomes a key task. Previously, determining the optimal length of data series for forecasting was a challenging task. Nowadays, with the use of modern technologies, this problem has been successfully solved. In this article, the authors propose an adaptive forecasting method to determine the optimal length of data series. The research results include forecasting error values with and without adaptation. The adaptive forecasting method demonstrated smaller values of mean absolute error (MAE) compared to the non-adaptive forecasting method (where the length of the time series is always equal to 100). This allows for more efficient and accurate prediction of cumulative disturbances, which is critically important for ensuring high and stable quality of asphalt concrete mixture.

Keywords: asphalt concrete, asphalt concrete mixture, disturbance, control system, autoregressive model, forecasting, adaptive forecasting method, optimal length of series, forecast accuracy, mean absolute error

Prompt adjustment of the composition of the asphalt concrete mixture is key to achieving high quality asphalt concrete. To enable easy and rapid adjustment of the asphalt concrete mixture formulation, predicting the properties of asphalt concrete (Marshall stability) is critically important. There are many methods for predicting the properties of asphalt concrete, but the choice of one method or another is a very pressing problem. This article proposes two methods for forecasting Marshall stability: forecasting using a multiple linear regression model and forecasting using an autoregressive model. To evaluate the forecasting accuracy of models, we use two metrics: average absolute error (MAE) and average absolute percentage error (MAPE). The results of the study show that the autoregressive model exhibits better forecasting results, especially the second-order autoregressive model.

Keywords: asphalt concrete, control, composition adjustment, forecasting, multiple linear regression model, autoregression model, Marshall stability, forecast accuracy, mean absolute error, mean absolute percentage error