Performance Improvement Of DAS In High-concentration Gas Detection And Research On CH4 Productization

With industry 4.0,peak carbon dioxide emissions and carbon neutrality accelerating,and industrial danger closing,it becomes more and more important to protect people’s lives and property.Most of the exhaust gas generated by industrial production is toxic and harmful,even flammable and explosive.In daily life,the demand for natural gas is gradually increasing.Therefore,the effective gas detection has become the best way to prevent explosions,reduce casualties,and protect people’s property.With the rapid development of laser technology,optical fiber sensing technology has gradually emerged in the field of gas detection with its high transmission speed,high sensitivity,safety and explosion-proof in recent decades.Especially,Tunable Diode Laser Absorption Spectroscopy Technology(TDLAS)has come the head of gas sensing with the advantages of non-contact,high precision,no sample consumption,and good real-time performance in the infrared band.A lot of exploration is worked in trace amount,high precision and high resolution at home and abroad for the gas detection industry.In practical applications,TDLAS not only is considered that the lower limit of the detection concentration is sufficiently low,but also that the upper limit of the detection concentration is sufficiently high.In other words,it is necessary to detect high concentrations and large-scale gas concentrations accurately.For example,the concentration of certain gases ranges is from parts per million(ppm)to 100%of the lower explosive limit(LEL)in tunnel engineering or mining operations.Specially,it is considered to corrected the fitting nonlinearity for detecting high concentrations of gases.Moreover,when high-concentration gases are detected by Direct Absorption Spectroscopy(DAS),absorption saturation will inevitably occur,overflowing the linear region,and even optical thickness(absorbance)bottoming.Therefore,it is imperative to research on gas detection methods and products.In this paper,firstly,a complete demodulation signal is obtained by dual-logarithmic demodulation method from the transmission signal.The fitting nonlinearity is eliminated.And calibrating the full concentration range are realized directly.Secondly,an artificial neural network model is applied in transmission spectroscopy to achieve gas concentration demodulation,which is a way of demodulating the concentration using multiple data points.Finally,the methane detection product is designed as an example,and the relevant performance indicators are tested.The main research contents are as follows:1.For detecting high concentration,dual-logarithmic demodulation method is proposed,correcting the nonlinear fitting problem successfully.The theoretical dual-logarithmic demodulation is expounded.And it is proven effective by high-concentration acetylene and the specific parameters and performance of the system are introduced in detail.Using double beams to balance the non-absorbing baseline,the absorption signal of this demodulation method is doubled,while the noise remains unchanged.The relative error between the actual demodulation peak and the simulated absorption peak is only 1.18%.At 1 atm,300 K,the optical thickness ranged from 0.0252 to 2.5335(L=300 cm,80-8100 ppm)maintain the good linearity(R2=0.9989).2.The CNN-BP model is constructed based on the convolutional neural network and the fully connected BP network.And the simulated data is used as the training sample to achieve a large range of concentration inversion successfully.Especially,to solve the problem that the absorption of high-concentration gas is completely saturated and the peak cannot be demodulated after optical thickness bottoming out.Finally,the relative error is less than 2%when the length of the gas chamber is 300 cm when the CH4 concentration is greater than 5000 ppm3.Productization is realized to detect methane,including power conversion module,temperature control and drive module,STM32 as the embedded peripheral module of MCU,concentration is inversed after acquiring transmission data,normalizing,filtering.Parameter debugging and performance test were carried out on the product prototype.When the ambient temperature changes from-8 to 50℃,the laser temperature change 0.3785℃,and the maximum wavelength drift is only about 31 pm;when the laser temperature changes by 1.63℃,the normalized peak value remains approximately unchanged.The maximum relative error is only 1.4%.When the ambient temperature is changed from-20 to 60℃,the normalized peak error corresponding to the same gas concentration is reduced from 61%to 7.5%by corrected;A total of 12 channels of 2 sets of different equipment are calibrated,getting an excellent set to invers concentration.And the normalized peak has a significant linear relationship with the concentration(R2=0.9998);In the last 48-hour long-term stability test,the maximum detection error of CH4 at 10000 ppm is only 3.12%.