A fitting frequency locking method, product, medium and device for cavity enhanced Raman spectroscopy

Abstract

The present application discloses a fitting frequency locking method, product, medium and device for cavity enhanced Raman spectroscopy, which relates to the field of semiconductor laser Raman spectroscopy detection technology. The method comprises: obtaining the error signal and transmission power of the resonant cavity in the cavity enhanced Raman spectroscopy detection system; determining effective features based on the error signal and transmission power; inputting the effective features into a trained back propagation neural network to obtain a fitting error signal; and fitting frequency locking the cavity enhanced Raman spectroscopy according to the fitting error signal. The fitting frequency locking method, product, medium and device for cavity enhanced Raman spectroscopy of the present application can effectively improve the frequency stability and anti-interference ability of the frequency locking system while ensuring the high sensitivity of the error signal.

Description

Technical Field

[0001] The present application relates to the field of semiconductor laser Raman spectroscopy detection technology, and in particular to a fitting frequency-locking method, product, medium and equipment for cavity-enhanced Raman spectroscopy. Background Art

[0002] Gas detection is of great significance in the fields of environmental monitoring, industrial processes and safety, medical diagnosis, etc., and with the progress and continuous improvement of science and technology, cavity-enhanced Raman spectroscopy technology, as an advanced and powerful gas detection technology, has gradually attracted widespread attention and currently plays an important role in the field of gas detection. Cavity-enhanced Raman spectroscopy technology greatly enhances the Raman signal of gas through the resonance enhancement effect of the optical resonant cavity. Compared with traditional ordinary Raman spectroscopy technology, cavity-enhanced Raman spectroscopy technology has higher sensitivity and specificity, and can achieve accurate detection and analysis of low-concentration gases. Cavity enhancement requires stability between the laser frequency and the cavity resonance frequency to ensure that the resonance enhancement effect in the optical resonant cavity is maximized. However, in the process of achieving optimal resonance enhancement, that is, maximizing the resonance enhancement effect, disturbances in the external environment and mechanical vibrations can easily lead to the loss of lock of the optical resonant cavity. Since the optical resonant cavity plays a vital role in cavity enhancement, it is of vital significance to ensure the stability and anti-interference ability of the optical resonant cavity for cavity-enhanced Raman spectroscopy technology.

[0003] Over the years, researchers have conducted in-depth research on different methods of laser frequency locking. PDH (Pound-Drever-Hall) technology has become a standard method for stabilizing optical resonators due to its advantages of high precision, automated control, fast response, and real-time monitoring, and is widely used in technical fields such as space laser interferometer gravitational wave detection, optical communications, optical atomic clocks, and laser frequency locking. Although PDH technology is currently the most widely used laser frequency locking method, it still has many defects. First, the linear dynamic range of the error signal of the PDH technology is very small for the high-precision FP (Fabry-Perot) resonator, making it a challenging task to control the servo system for frequency locking, and it is very susceptible to external interference or environmental changes, resulting in loss of lock; secondly, at frequencies other than the locking frequency of the optical resonator, there is a region where the error is zero and linear dynamic, so after being disturbed, it is easy to lock the frequency to an adjacent low resonant frequency, resulting in low or no resonance in the optical resonator, and ultimately leading to the disappearance of the Raman scattering signal.

[0004] In response to the above problems, scholars have proposed a variety of effective methods to improve the anti-interference ability of PDH frequency locking technology. Shinji Miyoki et al. from the University of Tokyo proposed two methods for the control of FP resonant cavity 8: near-Q phase demodulation and the use of multiple sets of in-phase odd harmonic modulation frequency demodulation to expand the linear dynamic range of the error signal of PDH technology. However, this method cannot guarantee the sensitivity of the error signal and reduces the transmission power of FP resonant cavity 8. Han Sen, Li Chan et al. from the University of Shanghai for Science and Technology divided the error signal of traditional PDH technology by the transmission power and redefined a PDH error signal. Under the condition of a resonant cavity length of 10m, the linear dynamic range of the frequency locking system was increased by 50 times. However, this method has a poor effect on increasing the linear dynamic range when the resonant cavity is short and greatly reduces the sensitivity of the error signal. Yan Liping et al. of Zhejiang University of Technology used a new error signal constructed under a large modulation depth for pre-locking, and then used the traditional error signal corresponding to a small modulation depth to achieve precise locking, which expanded the linear dynamic range of the frequency-locked system by 3.4 times and ensured the sensitivity of the error signal during locking. Although this method avoids the influence on the sensitivity of the error signal when the dynamic range is expanded, the large modulation depth will cause the nonlinear distortion of the signal to become significant and the increase in the linear dynamic range is limited. Manuel Zeyen et al. of the Institute of Particle Physics and Astrophysics in Germany phase-modulated the seed laser so that the modulation frequency was half of the free spectral range (FSR) of the FP resonant cavity 8, thereby eliminating the stable but non-resonant locking points in the error signal of the traditional PDH technology, also known as Trojan operation points. In this way, even after a large disturbance, the FP resonant cavity 8 can still correctly lock to the seed frequency. Although this method will not lock at the Trojan operation point, because the locked frequency range is narrow, the frequency-locked system is more susceptible to external interference, resulting in unstable locking or frequent frequency hopping.

[0005] In summary, the existing related frequency locking methods for solving the defects of PDH technology still cannot improve the frequency stability and anti-interference ability of the frequency locking system while ensuring the high sensitivity of the error signal. Summary of the invention

[0006] The purpose of this application is to provide a fitting frequency locking method, product, medium and equipment for cavity enhanced Raman spectroscopy, which can effectively improve the frequency stability and anti-interference ability of the frequency locking system while ensuring high sensitivity of the error signal.

[0007] To achieve the above-mentioned objectives, the present application provides the following solutions: In the first aspect, the present application provides a fitting frequency locking method for cavity-enhanced Raman spectroscopy, including: obtaining an error signal and a transmission power of a resonant cavity in a cavity-enhanced Raman spectroscopy detection system; determining effective features based on the error signal and the transmission power; inputting the effective features into a trained back-propagation neural network to obtain a fitting error signal; and performing fitting frequency locking of the cavity-enhanced Raman spectroscopy according to the fitting error signal.

[0008] Optionally, the step of acquiring the error signal and transmission power of the resonant cavity in the cavity enhanced Raman spectroscopy detection system specifically includes: performing piezoelectric scanning on the resonant cavity in the cavity enhanced Raman spectroscopy detection system, using a high-sensitivity sensor to collect the error signal of the resonant cavity in each scanning cycle, and synchronously using a photodetector to measure the transmission power of the resonant cavity in each scanning cycle.

[0009] Optionally, determining the effective feature based on the error signal and the transmission power specifically includes: determining the effective feature based on the error signal and transmitted power , the error signal The amplitude and slope, transmission power The amplitude, slope and signal ratio / The amplitude and slope of Regularization method selects features and converts the error signal The amplitude and transmission power The amplitude and signal ratio / The amplitude and slope of are taken as effective features.

[0010] Optionally, the trained back-propagation neural network is trained based on the Adam optimization algorithm with the goal of minimizing the loss function.

[0011] Optionally, the loss function of the back propagation neural network for training is the mean square error MSE: ; Where n is the total number of sample groups; i is the i-th group of samples; is the actual error signal of the i-th group of samples; is the fitting error signal of the i-th group of samples.

[0012] Optionally, the back propagation neural network includes an input layer, three hidden layers and an output layer; wherein the three hidden layers select 10, 8 and 5 nodes respectively.

[0013] Optionally, performing frequency locking of cavity enhanced Raman spectroscopy according to the fitting error signal specifically includes: when the fitting error signal The absolute value of is less than the first preset error threshold When the cavity enhanced Raman spectroscopy detection system automatically switches to the locked state and uses the fitting error signal ; When the fitting error signal The absolute value of is less than the second preset error threshold When the first preset error threshold Greater than a second preset error threshold , the CERAS system automatically switches to a locked state and uses the error signal ; If there is external interference, restore the fitting error signal .

[0014] In a second aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the fitting frequency locking method of the cavity enhanced Raman spectroscopy.

[0015] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the fitting frequency-locking method of the cavity-enhanced Raman spectroscopy.

[0016] In a fourth aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the fitting frequency-locking method of the cavity-enhanced Raman spectroscopy.

[0017] According to the specific embodiments provided by the present application, the present application discloses the following technical effects: the present application provides a method, product, medium and equipment for fitting frequency locking of cavity enhanced Raman spectroscopy. In order to further improve the stability and anti-interference ability of the resonant cavity, the present application mainly provides a wide range fitting frequency locking (WRFFL) method, that is, a method for fitting frequency locking of cavity enhanced Raman spectroscopy, by detecting the error signal and transmission power of the resonant cavity, combined with the back propagation neural network (BPNN), referred to as BP neural network for modeling, thereby obtaining a wide range, Trojan-free operation point fitting error signal, expanding the linear dynamic range of the error signal, and finally realizing an optical resonant cavity with high stability and strong anti-interference ability. In addition, the present application combines the cavity enhanced Raman spectroscopy detection system to test the frequency locking performance of the proposed WRFFL method, proving that the WRFFL method of the present application not only improves the anti-interference ability of the frequency locking system, but also greatly improves the frequency stability of the frequency locking system, expands the linear dynamic range of the error signal, and prolongs the frequency locking time of the resonant cavity while retaining the high sensitivity characteristics of the error signal. Therefore, the fitting frequency-locking method, product, medium and device of the cavity-enhanced Raman spectroscopy of the present application can effectively improve the frequency stability and anti-interference ability of the frequency-locking system while ensuring the high sensitivity of the error signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 A schematic diagram of the flow of the fitting frequency-locking method for cavity-enhanced Raman spectroscopy provided in this application.

[0020] Figure 2 This is a schematic diagram of the structure of the cavity enhanced Raman spectroscopy detection system provided in this application.

[0021] Figure 3 Schematic diagram of the principle of the fitting frequency-locking method for cavity-enhanced Raman spectroscopy provided in this application.

[0022] Figure 4 This is a simulation diagram of the error signal based on theory provided by this application.

[0023] Figure 5 Schematic diagram of multi-scan cycle error signals and Trojan operation points provided in this application.

[0024] Figure 6 A schematic diagram of the training process of the back-propagation neural network based on the Adam optimization algorithm provided in this application.

[0025] Figure 7 This is a schematic diagram of the fitting error signal curve and preset error threshold provided in this application.

[0026] Figure 8 A schematic diagram of a fitting error signal curve and a preset linear signal obtained by a trained back-propagation neural network provided in this application.

[0027] Fig. 9 This is a schematic diagram of the anti-interference test results provided for this application.

[0028] Fig.10 This is a schematic diagram of the error signal and reflected power changes over time provided by this application.

[0029] Figure numerals: spectrometer—5, dichroic mirror DM—6, polarization beam splitter PBS3—7, FP resonant cavity—8, piezoelectric ceramic PZT—9, photodetector PD1—10, mode matching lens PML—11, quarter wave plate—12, polarization beam splitter PBS2—13, piezoelectric ceramic controller—14, laser—15, current controller—16, digital-to-analog converter DAC—17, Faraday isolator FI—18, half wave plate—19, polarization beam splitter PBS1—20, electro-optic modulator EOM—21, radio frequency terminal RF—22, photodetector PD2—23, mixer Mixer—24, low-pass filter LPF—25, phase shifter—26, local oscillator LO—27, host computer—28, PID proportional integral differential feedback amplifier—29. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0031] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0032] The fitting frequency locking method of cavity enhanced Raman spectroscopy provided in this application is suitable for Raman spectroscopy detection technology with high stability and strong anti-interference ability. By constructing a cavity enhanced Raman spectroscopy detection system, the transmission power and modulation depth of the resonant cavity, that is, the FP cavity, are detected. The error signal is 1.08, and the linear dynamic range of the error signal is the narrowest and the sensitivity is the highest at this modulation depth. Subsequently, through a three-layer back-propagation neural network, a wide range linear fitting of the error signal is achieved based on multiple sets of eigenvalue data of effective features determined by the error signal and the transmitted power, providing a wide range and Trojan-free operation point fitting error signal for the PDH frequency locking system, and finally realizing a cavity-enhanced Raman spectroscopy detection system with high frequency stability and strong anti-interference ability.

[0033] like Figure 2 As shown, the present application uses optical elements and electronic components to build an optical path and electrical detection structure unit, and debugs the optical path and photoelectric detection system, forming the main structure of a cavity-enhanced Raman spectroscopy detection system with wide-range fitting and frequency locking. The cavity-enhanced Raman spectroscopy detection system is mainly composed of an optical path structure unit and a circuit structure unit.

[0034] In the optical path structure unit, the present application uses a 532nm solid-state single longitudinal mode laser 15 as a pump light source. The spectral line width of the output light emitted by the light source is less than 1MHz. Through the Faraday isolator FI18, the Faraday isolator FI18 can reduce the reflection efficiency, thereby preventing the laser 15 from being damaged by laser reflection. The waist position of the output light of the laser 15 is measured by the knife-edge method at a distance of 25cm from the output port of the laser 15, and the waist size is 0.72mm. Since the polarization direction required for the incident light of the electro-optic modulator EOM21 is horizontal polarization, a polarizer is added between the laser 15 and the electro-optic modulator EOM 21, that is, the output laser is converted into p-light as much as possible through a 1-inch half-wave plate HWP (half-wave plate 19), also known as TM polarized light, and then the remaining s-light is passed through the polarization beam splitter PBS1 20, allowing the p-light to enter the electro-optic modulator EOM 21. The electro-optic modulator EOM 21 can provide an adjustable phase difference for linearly polarized incident light. The role of the electro-optic crystal material inside the electro-optic modulator EOM 21 in the Raman spectrum is to cause a phase difference in the light, thereby enhancing the detection effect of the Raman signal. Therefore, the electro-optic modulator EOM 21 changes the phase of the light so that the phase difference of the light changes after modulation, thereby improving the signal-to-noise ratio and detection sensitivity of the Raman signal. A voltage is applied through the radio frequency end RF 22 of the electro-optic modulator EOM 21 to change the extraordinary light refractive index of the electro-optic crystal inside the electro-optic modulator EOM 21, so that the optical signal produces a phase difference. This change in phase difference can help separate the Raman signal from the background noise, improve the contrast of the signal, and thus enhance the detection effect of the Raman signal. The input signal of the radio frequency end RF 22 is a 20MHz signal generated by the local oscillator LO 27. Then the output signal of the electro-optic modulator EOM 21 passes through the polarization beam splitter PBS2 13, and the linear polarized light is converted into circularly polarized light through the quarter-wave plate 12. The laser is then accurately coupled to the FP resonant cavity 8 through the mode matching lens PML 11. The FP resonant cavity 8 is composed of two high-reflectivity lenses, so that the laser entering the FP resonant cavity 8 is divided into three parts: one part is transmitted through the polarization beam splitter PBS3 7 and then enters the photodetector PD1 10. The photodetector PD1 10 sends the collected transmission power of the FP resonant cavity 8 to the host computer 28; another part passes through the polarization beam splitter PBS3 7 and then passes through the dichroic mirror DM 6 to enter the spectrometer 5; the last part returns along the original route and passes through the quarter-wave plate 12 to be converted from circularly polarized light to linearly polarized light. At this time, the polarization state has been rotated 90° relative to the incident laser, so it will enter the photodetector PD2 23 after passing through the polarization beam splitter PBS2 13, and then enter the circuit structure unit.

[0035] In the circuit structure unit, the signal obtained by the photodetector PD2 23 is first demodulated, so it first passes through a mixer Mixer 24. Another signal input to the mixer Mixer 24 is the same 20MHz signal generated by the local oscillator LO 27 before, which is phase-shifted by 90° by the phase shifter 26. In this way, the output signal of the photodetector PD2 23 and the phase-shifted signal of the local oscillator LO 27 enter the mixer Mixer 24 and pass through the low-pass filter LPF 25 to filter out the high-frequency signal, and obtain the error signal of the PDH frequency-locked system. , and at the same time, the transmitted power of the FP resonant cavity 8 is measured using the photodetector PD1 10 and synchronized with the error signal The error signal After filtering, the detected value at each sampling point is The amplitude of The amplitude and signal ratio / The four characteristic values ​​of the amplitude and slope are used as the input data of the back propagation neural network for training to obtain the fitting error signal, and the fitting error signal is input into the PID proportional integral differential feedback amplifier 29, compared with the error signal obtained by the low-pass filter LPF 25, and a feedback signal is output. After the feedback signal of the PID proportional integral differential feedback amplifier 29 is processed by the digital-to-analog converter DAC 17, a fast feedback signal is output to the current controller 16 port of the laser 15, and another slow feedback signal is output to the modulation port of the piezoelectric ceramic controller 14 on the high-reflection mirror of the FP resonant cavity 8, and the piezoelectric ceramic PZT 9 is controlled to adjust the cavity length of the FP resonant cavity 8 according to the input slow feedback signal, so that the cavity length matches the wavelength of the laser 15, and the mode locking of the FP resonant cavity 8 and the laser 15 is realized.

[0036] To avoid confusing the essence of the present application, the known components and circuits are not described in detail. To better understand the technical solution of the present application, the following will be further described in combination with the theoretical calculation of the present application.

[0037] Laser 15 emits light signal It can be expressed as ,in is the light field intensity, is the angular frequency, t is the time, and a frequency of is applied to the electrode of the electro-optic modulator EOM 21 through the input of the radio frequency terminal RF 22. The modulation signal has a modulation depth of , the modulated laser Become the following formula.

[0038] (1).

[0039] when 1 o'clock, It can be transformed into the following formula.

[0040] (2).

[0041] in, , denote zero-order and first-order Bessel functions, respectively.

[0042] At this time, the laser contains a carrier and two angular frequencies respectively. The signal then enters the FP resonant cavity 8. Since the error signal size is determined by detecting the reflected signal, the transfer characteristics of the FP resonant cavity 8 need to be obtained. is the angular frequency of the incident light Resonance angular frequency The frequency difference function between them, assuming that the FP resonant cavity 8, referred to as the resonant cavity, is a symmetrical lossless cavity, then the frequency difference function Is the following formula.

[0043] (3).

[0044] in, is the reflected light signal; is the reflectivity of the high-reflectivity mirror of the resonant cavity; is the free spectral range of the resonant cavity.

[0045] The reflected light signal can be obtained by formula (2) and formula (3): The complex amplitude electric field representation of is as follows.

[0046] (4).

[0047] in, To apply an angular frequency of The modulus of the frequency difference function after the modulation signal.

[0048] The reflected light intensity As shown below.

[0049] (5).

[0050] in, , is the effective power factor. Indicates the conjugate operation.

[0051] In formula (5), the DC signal and high-frequency signal are removed. , the rest is the intermediate frequency signal, which is amplitude modulated As shown below.

[0052] (6).

[0053] Formula (5) is used to filter out the DC and high-frequency components, and combined with formula (6), the following formula (7) is obtained.

[0054] (7).

[0055] At this time, the That is, the error signal required by this application is derived based on theoretical principles and calculated based on theoretical basis.

[0056] It can be seen that there are two error signals in formula (7), but only one of them is valid for different modulation frequencies. The modulation mode is divided into high-frequency modulation and low-frequency modulation. When the modulation frequency is much smaller than the line width of the resonant cavity, it is low-frequency modulation, otherwise it is high-frequency modulation. In practical applications, high-frequency modulation is generally used, such as Figure 4 As shown in (a), Figure 4 (a) is an error signal diagram based on theory for high-frequency modulation. Since high-frequency modulation has a higher frequency response, high signal-to-noise ratio, high measurement accuracy, and better anti-interference ability, it effectively ensures the frequency stability in optical frequency transmission.

[0057] The two most important indicators of the error signal are the linear dynamic range and sensitivity, which directly affect the effect and accuracy of the frequency lock. The linear dynamic range refers to the frequency axis range between the maximum amplitude and the minimum amplitude of the error signal near the zero point. The slope of the curve within the linear dynamic range is defined as the sensitivity of the error signal. , since formula (7) is continuously differentiable, the sensitivity can be determined by taking the derivative of the error signal with respect to the frequency error The exact value of Figure 4 As shown in (b), Figure 4 (b) is the differential diagram of the error signal versus frequency. Sensitivity has an important influence on the stability, frequency noise, and noise floor of the frequency-locking system. When the frequency offset is less than the half-height full width of the etalon spectrum, It can be simplified as follows.

[0058] (8).

[0059] Among them, the quality factor , then the sensitivity as follows.

[0060] (9).

[0061] From formula (9), we can see that when the incident power, cavity quality factor and free spectral range remain unchanged, the sensitivity Only with modulation depth Regarding, according to Figure 4 As shown in (c) Follow The change curve diagram can be obtained when hour reaches its maximum value, at which point the sensitivity of the error signal maximum.

[0062] Figure 5 Schematic diagram of error signal and Trojan operation point in multiple scanning cycles. The linear dynamic range of the error signal center is relative to one cycle, i.e., the free spectrum range of the resonant cavity. is very narrow, and Frequency locking requires locking the error signal at the position where the amplitude is zero in the linear signal. At this time, the transmission power of the resonant cavity is the largest, that is, the generated Raman signal is the strongest. However, a tiny external disturbance may cause the system to lose lock, resulting in a huge attenuation of the cavity transmission signal and the inability to detect the Raman signal. In the process of relocking after losing lock, in addition to the error signal point with the maximum transmission power in one cycle, there are also two points with low transmission power and two Trojan operation points with no transmission power. Therefore, expanding the linear dynamic range of the error signal and eliminating the Trojan operation point are the key to ensuring the high stability and strong anti-interference ability of the cavity-enhanced Raman spectroscopy detection system.

[0063] In order to solve the problems of narrow linear dynamic range and Trojan operation point of the above-mentioned error signal, the present application uses a back propagation neural network to perform linear fitting on the error signal. To ensure the three conditions of high sensitivity, wide dynamic range and no Trojan operation point, the back propagation neural network will perform multi-segment, wide-range linear fitting.

[0064] For the back propagation neural network, we first establish the back propagation neural network structure, which includes an input layer, three hidden layers and an output layer. The three hidden layers have 10, 8 and 5 nodes respectively. Since linear fitting is relatively simple, keeping the hidden layer small can avoid overfitting. Function as the activation function of the back-propagation neural network, yes An improved version of , which retains a small slope in the negative part, helps solve the problem of "neuron death". , where x is the eigenvalue of the effective feature of the back propagation neural network input. Specifically, the input layer is the first layer of the back propagation neural network, which is used to receive external input data. Each node in the hidden layer usually represents a feature, and the number of nodes is determined by the feature dimension of the input data. The output layer is the last layer of the back propagation neural network, which is used to generate the prediction result, that is, the fitting error signal.

[0065] like Figure 1 and Figure 3 As shown, the fitting frequency-locking method of cavity-enhanced Raman spectroscopy disclosed in the present application specifically includes the following steps.

[0066] Step 1: Obtain the error signal and transmission power of the resonant cavity in the cavity-enhanced Raman spectroscopy detection system.

[0067] Since the fitting frequency locking method of cavity enhanced Raman spectroscopy in this application requires a trained back propagation neural network, in order to obtain the training and verification data sets required in the back propagation neural network, combined with Figure 2 , piezoelectric scanning is performed on the resonant cavity in the cavity enhanced Raman spectroscopy detection system. The output power of the laser 15 is fixed at 50mW, the reflectivity of the two high-reflection mirrors is 99.9%, the phase shift is fixed at 90°, and the modulation frequency is 20MHz. First, a triangular wave scanning signal is applied to the laser 15 or the piezoelectric ceramic in the resonant cavity, and the length of the resonant cavity or the laser frequency is linearly scanned. The scanning distance is adjusted so that there is only a single fundamental mode signal in the scanning cycle. After matching, a high-sensitivity photoelectric sensor, that is, the photodetector PD2 23 is used to record the error signal in each scanning cycle to obtain the modulation depth The error signal is 1.08 , and the transmitted power of the resonant cavity is measured using the photodetector PD1 10 And synchronized with the collection of error signals, multiple groups of error signals and transmission powers are obtained to form a sample set for subsequent training and verification of the back-propagation neural network.

[0068] Step 2: Determine the effective features based on the error signal and the transmitted power.

[0069] In order to select the most effective feature points for input into the back propagation neural network, this application is based on the error signal and transmitted power , the error signal The amplitude and slope, transmission power The amplitude, slope and signal ratio / The amplitude and slope of are used as features. However, since too many features will increase the computational complexity, cause the gradient to disappear or explode, and make the model complicated, this application adopts Regularization method selects features and converts the error signal The amplitude and transmission power The amplitude and signal ratio / The amplitude and slope of are taken as effective features.

[0070] The regularized result contains the fitted coefficient matrix And fitting information Two important parameters. The coefficient matrix It can provide the degree of contribution of the feature in the model, and its absolute value can be used as the basis for feature selection. Usually, a screening threshold can be set to select important features.

[0071] use The specific steps of the regularization method to screen the above features are as follows.

[0072] 1. Prepare a data set containing features and perform necessary digital filtering on the data.

[0073] 2. Use the constructed back propagation neural network and The loss function of the regularization term constructs a data training model to train the data.

[0074] 3. After training is completed, obtain the coefficient matrix of the data training model .

[0075] 4. According to the coefficient matrix The size of the coefficient is used to screen out the eigenvalues ​​of effective features.

[0076] As a specific embodiment, the present application adopts Regularization screening, selection of error signal The amplitude and transmission power The amplitude and signal ratio / The amplitude and slope of Specifically, the present application filters the error signal and converts the error signal detected at each sampling point into , Transmitted power and its ratio / The characteristic values ​​of the amplitude and slope of the four features are used as a set of input data of the back propagation neural network, and the target linear signal generated in real time by theoretical simulation at the sampling point is used as the ideal preset linear signal, such as Figure 8 As shown, this signal is used as the response data to train and verify the back propagation neural network, and the mean square error between the predicted value and the true value is used as the loss function of the training.

[0077] Step 3: Input the effective features into the trained back-propagation neural network to obtain the fitting error signal.

[0078] Mean square error ) is used as the loss function for back-propagation neural network training, that is, the average of the squared differences between the predicted values ​​and the true values ​​of the back-propagation neural network, as shown in the following formula.

[0079] (10).

[0080] Where n is the total number of sample groups; i is the i-th group of samples; is the actual error signal (true value) of the i-th group of samples; is the fitting error signal (predicted value) of the i-th group of samples.

[0081] In this application, effective features are input into a trained back-propagation neural network, and the trained back-propagation neural network is linearly fitted to generate a fitting error curve, such as Figure 7 As shown by the slant line in , the fitting error point can be obtained from the fitting slant line.

[0082] In addition, although the above-mentioned back propagation neural network has strong generalization ability, approximation ability and adaptive ability, it is easy to fall into the problem of local optimal solution. The optimization algorithm optimizes the weights and thresholds in the back propagation neural network. The back propagation neural network optimized by adaptive moment estimation overcomes the problem of local optimal solution, and has fast convergence speed and good robustness. The optimizer defines the algorithm. The back propagation neural network of the optimization algorithm builds a model between the error signal and the target error signal, where The algorithm flow chart is as follows Figure 6 shown.

[0083] right The optimizer defines the algorithm, and its algorithm steps are as follows.

[0084] 1) First initialize the parameters: learning rate Set to 0.001, which determines the step size of each parameter update, momentum Set to 0.99, in order to accumulate historical gradient information, help accelerate convergence, and better guide the update direction of parameters, the second moment estimate Set to 1. Choosing 1 can make the second moment estimate quickly adapt to the changes in the gradient, thereby better adjusting the learning rate.

[0085] 2) Calculate the gradient of the loss function using the current parameters; update the momentum , accumulates the exponentially weighted average of the current gradient into momentum.

[0086] 3) Update the second moment estimate , accumulating an exponentially weighted average of the squares of the current gradients into the second moment estimate.

[0087] 4) Calibration , The deviation is corrected using , To update the parameters .

[0088] 5) Repeat the above iterations.

[0089] The mean square error threshold is set to , the maximum number of iterations is set to 50.

[0090] Step 4: Perform fitting frequency locking of the cavity enhanced Raman spectroscopy based on the fitting error signal.

[0091] The fitting error signal is obtained through the trained back propagation neural network (ADAM-BP) The fitting error curve is as follows: Figure 8 As shown, the fitting error signal is obtained from the fitting error curve and a first preset error threshold and a second preset error threshold , the parameter values ​​are as follows Figure 7 As shown. When the fitting error signal The absolute value of is less than the first preset error threshold When the cavity enhanced Raman spectroscopy detection system automatically switches to the locked state and uses the fitting error signal ; When the fitting error signal The absolute value of is less than the second preset error threshold When the first preset error threshold Greater than a second preset error threshold , the CER detection system automatically switches to a locked state and uses a more sensitive error signal If there is external interference, the transmitted power drops to the second power threshold Below, the fitted error signal is recovered In addition, if the transmitted power is lower than the first power threshold , fitting error signal Exceeding the first preset error threshold , the cavity enhanced Raman spectroscopy detection system will start rescanning. The fitting error signal expands the interval and range of the error signal. The first preset error threshold here is And the first power threshold It is obtained by averaging the corresponding values ​​obtained by repeated tests when the target linear signal responds at the corresponding sampling point. Figure 7 As shown in the figure, each piezoelectric displacement corresponds to a transmission power line composed of different transmission power values. Since the selection of piezoelectric displacement is relatively dense, multiple transmission power lines are densely arranged to form Figure 7 The area shown in the lower half.

[0092] In summary, if Figure 3 As shown, the principle of the fitting frequency locking method of cavity enhanced Raman spectroscopy proposed in this application mainly includes an error signal generation module, a back propagation neural network training and output module, and an error signal selection and control output module. The locking process of this application includes two steps: scanning and locking. First, a triangular wave scanning signal is applied to the piezoelectric ceramic in the resonant cavity or the laser 15, and the cavity length or laser frequency is linearly scanned. After matching, the error signal is obtained. , and according to the transmitted power get / The amplitude and slope are used to obtain the fitting error signal through the trained back propagation neural network , and the fitting frequency locking of the cavity enhanced Raman spectroscopy is performed according to the fitting error signal. In addition, the data DATA is a large amount of data sets of stored error signals and transmission power collected by the cavity enhanced Raman spectroscopy detection system as the input for training the back propagation neural network.

[0093] In addition, in order to characterize the anti-interference ability of the cavity-enhanced Raman spectroscopy detection system based on the fitting frequency-locking method of cavity-enhanced Raman spectroscopy proposed in this application, interference is applied to the traditional cavity-enhanced Raman spectroscopy detection system and the cavity-enhanced Raman spectroscopy detection system of this application, and the results are as follows: Fig. 9 As shown, Fig. 9 (a) is a traditional CER detection system, i.e., the change of error signal and piezoelectric ceramic control voltage after the original CER system is disturbed. After the disturbance is applied, the original CER system loses lock and cannot be re-locked. Will remain in , while the error signal is almost 0. Fig. 9 (b) is a diagram showing the change of the error signal and the piezoelectric ceramic control voltage after the cavity enhanced Raman spectroscopy detection system of the present application is disturbed. After the cavity enhanced Raman spectroscopy detection system of the present application is disturbed, the piezoelectric ceramic control voltage signal quickly changes from Back to The error signal recovers to its previous state after fluctuation, that is, the system returns to the locked state after a short detuning, indicating that the cavity-enhanced Raman spectroscopy detection system of the present application has better anti-interference ability.

[0094] In order to characterize the stability of the cavity enhanced Raman spectroscopy detection system based on the fitting frequency locking method of cavity enhanced Raman spectroscopy proposed in this application, frequency locking experiments were carried out using an error signal and a fitting error signal under the same conditions. After multiple experiments, it was found that the traditional cavity enhanced Raman spectroscopy detection system has a locking time of 15-20 minutes and is very easy to lose lock. After losing lock, the transmitted power is 0 and the reflected power is greatly enhanced. Fig.10 This is a schematic diagram of the error signal and reflected power change of the cavity enhanced Raman spectroscopy detection system of this application. Affected by ambient temperature fluctuations and environmental disturbances, the error signal amplitude always fluctuates around 0, and its maximum fluctuation is about , the reflected power amplitude also fluctuates around its minimum value, and its maximum fluctuation is about , and the frequency locking time is much longer than that of the traditional cavity-enhanced Raman spectroscopy detection system, and is stable for more than 200 minutes, indicating that the cavity-enhanced Raman spectroscopy detection system of the present application has good long-term stability.

[0095] In some embodiments, the present application also provides a computer program product, including a computer program, which implements the fitting frequency locking method of cavity enhanced Raman spectroscopy when executed by a processor.

[0096] In some embodiments, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the fitting frequency-locking method of cavity-enhanced Raman spectroscopy is implemented.

[0097] In some embodiments, the present application also provides a computer device, including a processor, a memory, an input / output interface (I / O for short), a communication interface, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the described cavity-enhanced Raman spectroscopy fitting frequency-locking method.

[0098] Wherein, the processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Wherein, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store pending transactions. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the fitting frequency locking method of the cavity enhanced Raman spectroscopy is implemented.

[0099] It should be noted that the object information (including but not limited to object device information, object personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the object or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards.

[0100] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0101] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0102] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A fitting frequency locking method for cavity enhanced Raman spectroscopy, characterized in that: include: Obtain the error signal and transmission power of the resonant cavity in the cavity enhanced Raman spectroscopy detection system; determining a valid feature based on the error signal and the transmitted power; Input the effective features into the trained back-propagation neural network to obtain the fitting error signal; Perform frequency locking of cavity enhanced Raman spectroscopy according to the fitting error signal; The determining of effective features based on the error signal and the transmitted power specifically includes: Based on the error signal and transmitted power , the error signal The amplitude and slope, transmission power The amplitude, slope and signal ratio / The amplitude and slope of are used as features; use Regularization method selects features and converts the error signal The amplitude and transmission power The amplitude and signal ratio / The amplitude and slope of are taken as effective features.

2. The fitting frequency locking method for cavity enhanced Raman spectroscopy according to claim 1, characterized in that: The step of obtaining the error signal and transmission power of the resonant cavity in the cavity enhanced Raman spectroscopy detection system specifically includes: The resonant cavity in the cavity-enhanced Raman spectroscopy detection system is piezoelectrically scanned, a high-sensitivity sensor is used to collect the error signal of the resonant cavity in each scanning cycle, and a photodetector is used synchronously to measure the transmission power of the resonant cavity in each scanning cycle.

3. The fitting frequency locking method for cavity enhanced Raman spectroscopy according to claim 1, characterized in that: The trained back propagation neural network is trained based on the Adam optimization algorithm with the goal of minimizing the loss function.

4. The fitting frequency locking method for cavity enhanced Raman spectroscopy according to claim 3, characterized in that: The loss function of the back propagation neural network training is the mean square error MSE: ; Where n is the total number of sample groups; i is the i-th group of samples; is the actual error signal of the i-th group of samples; is the fitting error signal of the i-th group of samples.

5. The fitting frequency locking method for cavity enhanced Raman spectroscopy according to claim 1, characterized in that: The back propagation neural network includes an input layer, three hidden layers and an output layer; wherein the three hidden layers select 10, 8 and 5 nodes respectively.

6. The fitting frequency locking method for cavity enhanced Raman spectroscopy according to claim 1, characterized in that: The fitting and frequency locking of the cavity enhanced Raman spectroscopy according to the fitting error signal specifically includes: When fitting the error signal The absolute value of is less than the first preset error threshold When the cavity enhanced Raman spectroscopy detection system automatically switches to the locked state and uses the fitting error signal ; When the fitting error signal The absolute value of is less than the second preset error threshold When the first preset error threshold Greater than a second preset error threshold , the CERAS system automatically switches to a locked state and uses the error signal ; If there is external interference, restore the fitting error signal .

7. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the fitting frequency-locking method for cavity-enhanced Raman spectroscopy according to any one of claims 1 to 6 is implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the fitting frequency-locking method for cavity-enhanced Raman spectroscopy according to any one of claims 1 to 6 is implemented.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the fitting frequency-locking method for cavity-enhanced Raman spectroscopy according to any one of claims 1 to 6.

Images