Problems of development of gas analysis system based on computing fabrics have been studied. The structure of the data gathering system that will be used for the natural gas composition analysis is proposed. The most significant features and the main advantages of the proposed system are described. The most appropriate statistical models that can be used in solving the task of natural gas composition analysis are presented. The consecutive stages for statistical model development and correlation analysis results are shown. Algorithms have been developed for effective solution for a number of gas analysis tasks for distributed systems. The mathematical basis of the learning algorithm and the architecture of the proposed neural network model are described. The different training completion cases for the developed neural network model are shown. Recommendations are made for effective application of neural network tools as the main models for solving the task of the natural gas composition analysis. Opportunities for further development of the proposed system are considered.