Drill Test Signal Denoising Analysis And Signal De-noising Method Based Threshold Learning

The development and researches of the petroleum bits are established on the theories and experiments, and a great deal of bit products of development and studying all need to pass indoor experiment and the spot experiment to complete. There are massive noises in the data due to test condition difference. In the past, noises which showed clearly in the data usually are eliminated artificially, or even not be disposal at all. Thus the result often can not affect the characteristic of the experiment, wasting a great deal of manpower and financial powers. Therefore, for acquiring dependable result in bits experiment, we need to study science of experiment data processing method urgently.Aim at above this actual engineering problem, the theory of wavelet transfer de-noise have been introduced. Wavelet has good localizing quality at time domain and frequency domain simultaneously and the characteristic of multi-resolution ratio analysis, so it can fulfill all kinds of wave-filtering needs such as low-pass, high-pass, sink wave, random noise de-noising. Compare with traditional wave-filtering methods, wavelet has incomparable advantage. Wavelet has become an effective means of signal analysis and is entitled as math microscope of signal analysis. This text commences with the bit gear rotation experiment and the measurement of tooth force experiment, analyzing their test principle and designing their test project in detail. According to test the signal characteristics with noise, we respectively adopt to package wavelet float the threshold de-noise method and wavelet de-noise method, and deal with bit gear rotation signal and measurement of tooth force signal, obtaining good result.In the wavelet de-noise calculation, the choice of the threshold is count for much. The existing threshold calculation ways are all according to the Donoho given nonlinear wavelet transfer threshold method. The selection of threshold all suppose noise is the Gauss noise, passing the statistical method to acquire. But for actual application systems may be the noise of other types, and making use of the Donoho method is not necessarily to get well de-noise result. Aim at this problem, this text presents a kind of thresholdself-study wavelet de-noise method. This method is applied in test systems that have determination and can carry the standard signal. This way pass to lead the nerve network nonlinear threshold unit and train the method into wavelet de-noise, at the same time making use of the nerve network to study the standard sample. So, it can ascertain threshold of by the square certain that time, being advantageous to looking for actual measurement system more excellent de-noise threshold.