Regression estimation of gas concentration in closed-loop control ventilation systems

The analysis of the gas concentration is an important problem in a wide area of industry and several home applications. It concern medicine, chemical industry, mountain, gas industry. The important significance such analysis has in building and construction engineering, housing et cetera. Now different systems are applied to a control of concentration, including the closed-loop ventilation schemes. In a lot of them sensors with high accuracy are applied. But, as a rule, applied effective techniques of estimation of concentration in a feedback control are rarely. In this paper presented new closed-loop scheme of monitoring and control of gas concentration. It uses performance, reliable approach of regression function evaluation that suits well for several of estimation problems and applications. In paper described application of the least-squares method for solving the estimation of concentration problem. Cases of nonlinear dependence regression function of time and estimation approaches for such cases considered. Means of obtaining the interval bounds for derived regression function are in details described. Both the cases of a linear dependence concentration of time and nonlinear dependence explained. Choices of nonlinear regression function linearization considered. Influence of confidence intervals on action of ventilation control system, its reliability and reaction under sudden perturbations investigated closely. Methods of definition upper and lower regression bounds depending on sensors instrumental errors are given. These bounds determine the prognosis of regression function and system behavior in general and used as a condition for closed-loop control. Implementations of the approach and simulation results are reported. Modeling of action the real system with this method presented. Comparison with the similar ventilation control schemes without applying regression estimation is given. Described method provides high reliability and robustness with computational simplicity using standard precision sensors and closed-loop clarity and minimalism.