Technological processes are always accompanied by deviations from the set mode, which is due to the influence of many external and internal factors. The environmental parameters, the components of input raw materials, and the condition of technological equipment are constantly changing, which requires solving the problem of finding the optimal control parameters and, in some cases, the parameters of the process itself.
 Most industries are focused on obtaining the final product with a given level of quality. Changes in parameters of the technological process may deteriorate the quality of production and cause defects or even emergency situations. To prevent this, forecasting methods are used.
 The task of constructing predictive models based on experimental data is relevant for a wide range of technological processes. Today, predictive models are widely used in management, diagnosis and identification. The vast majority of these models are based on artificial intelligence technologies or methods of mathematical statistics.
 The most widespread forecasting models find application in areas such as banking, insurance, business economics, medicine, diagnostics of technical components and equipment, and forecasting the parameters of technological processes.
 Despite the well-developed algorithm for model development and application, the main problem that remains is to acquire data, select an appropriate model structure, and integrate the model into existing control systems.
 The paper will predict the parameters of the technological process of methanol production under reduced pressure. The production of methanol under reduced pressure is a multi-stage process, and the emergence of problems at some stage will adversely affect further work and the end result.
 Note that there are all problems related to the performance of technological processes in the production of methanol, which are described above. Therefore, another task is to forecast emergencies, taking into account the indicators of all stages in the process. The development of models for forecasting emergencies and controlling thermal regimes and their further integration into the existing automatic process control system is proposed to be performed according to the principles of industrial revolution – Industry 4.0.
 Important components of Industry 4.0 are the Internet of Things, data analysis, and digital duplicates. Each of these components solves a partial problem and, collectively, they provide full automation of production, forecasting of real-time process indicators, and calculation of optimal control.
 The process of methanol production under reduced pressure can be fully automated in accordance with the components of Industry 4.0. First, there is instrumentation that allows the values of technological process to be obtained over time. Second, given a moderate size of these data, one can obtain models of control objects, perform their software implementation, and use them to calculate optimal control or predict the state of the process.
 The paper proposes a variant of constructing a virtual model based on experimental data and its further use with actual values of process parameters.
 A regression model was chosen to develop a model for predicting the temperature regime. Regression analysis allows checking the statistical significance of the parameters, assessing the adequacy and accuracy of the model, and establishing the nature and closeness of the relationship between the studied phenomena.
 It is also important to predict the occurrence of emergency (adverse) situations at the workplace. For this purpose, it is necessary to determine a list of these situations according to the technological regulations and develop a model for forecasting emergencies. There are various forms of presenting a model for forecasting emergencies. A decision tree is one of them. It will be developed for the production of methanol.
 The resulting tree is a graphical structure of the verbal (semantic) model relying on the expert's reasoning in solving problems related to emergencies. This is a network structure, whose nodes indicate potential deviations of the control object from the normal mode of operation. The resulting tree is used to solve forecasting and diagnosing problems.
 For practical use, the decision tree is implemented in software as an "if - then" set of rules. The software is used as an element of a higher-level system in relation to the existing automatic process control system.
Read full abstract