Numerical analysis of time series of the mine air parameters

Authors: J. Szywacz, S.  Wasilewski

Implemented in coal mines, the systems for monitoring and control of gas hazards and ventilation performance provide a great number of data onto the surface.

The purpose of those systems is current control maintained by the mine's dispatch service in relation to the mine atmosphere, and in particular in the aspects of potential methane and fire hazards.

As the test had proved (Dziurzyński, Wasilewski 1999), such data may be also applied for both the analyses and prophylactic-preventive measures carried out by ventilation services in mines. This paper presents is the analysis of time series of signals from measurement of the mine air, registered in systems for monitoring and control of ventilation.

Observation of the signals representing physical-chemical parameters of the mine air (pressure, air flow velocity or gases CH4, CO, and smoke concentration) prove that the above are subjected to disturbances of random amplitude and duration.

Reliability and effectiveness of the analyses being carried out require both the information on properties of physical-chemical parameters of mine air and identification of the signals' frequency/time structure.

Occurrence of the random disturbances in ventilation process and insufficient knowledge of the venue structure bring about a necessity of the statistic methods of identification to be applied in practice. A significant advantage of the statistic methods of processes identification (Mańczak 1971) may be the fact that their application docs not require experiments, but is based on data registered in conditions of the venue's regular operation.

These are just the most appropriate methods that enable to evaluate characteristics of the objects subjected to non-measurable random disturbances. For this purpose a numerical analysis of measuring data, based on the correlation-spectral theory of stationary stochastic processes is proposed.

The air physical-chemical parameters are the continuous realisation of the stochastic process observed only at discrete moment in time, because in computerised monitoring systems they are subjected to sampling with a fixed frequency, giving a sequence of real numbers that form a time series. As a result of the numerical analysis of a stochastic process the characteristics of signals (Bendat, Piersol 1976; Otnes, Enochson 1978) with regards to amplitude (statistical analysis), time (correlative analysis) and frequency (spectral analysis) were determined.

In the correlative-spectral analysis of time series the calculation algorithms based on the discrete Fourier transform are applied. Independently of individual signals testing, the paper presents the results of testing of inter-dependencies between signals registered at various points of the mine workings network.

So obtained information may be applied to the multi-aspect identification of the network parameters, with the use of signals correlation. The numerical analysis of time series in a complex network may be also used (Wasilewski 1998) for evaluation and balance of gases being released during of exploitation work, e.g. shooting, or gas-dynamic phenomena caused by rapid outflow of methane due to bump, methane breakout or explosion (Report of the Commission WUG, 2002).