A time domain double waveform combined urban underground space electromagnetic detection method

Abstract

The present application relates to a kind of city underground space electromagnetic detection methods of time domain double waveform combination, the purpose is to adopt trapezoidal-hemisine wave combination launch system, low-noise continuous acquisition system and neural network full waveform data interpretation method to carry out fine detection to city underground space geological structure.First, the detection ability of small loop transmitting coil is quantitatively analyzed based on the concept of cumulative sensitivity, and trapezoidal-hemisine wave combination launch system is used as excitation source;Then, double-channel low-noise continuous acquisition system is used to record full waveform synchronously on transmitting current and electromagnetic response;Finally, based on neural network algorithm, full waveform resistivity parameters are extracted from trapezoidal-hemisine wave measured response, and double waveform joint resistivity imaging is realized based on apparent resistivity weight coefficient.Through the application of time domain double waveform combination electromagnetic detection method, complex target body in city underground space can be effectively detected, and the detection range and accuracy of transient electromagnetic method to shallow abnormal body are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to a kind of urban underground space electromagnetic detection methods of time domain double waveform combination, it is suitable for electromagnetic geophysical exploration or geological structure detection field, especially to provide solution for urban underground space time domain electromagnetic fine detection. BACKGROUND

[0002] Magnetic source transient electromagnetic method (TEM) has the advantages of wide detection depth coverage and strong adaptability to working environment, and has been widely used in metal ore exploration, water resource survey, urban underground space structure detection and tunnel geological survey, etc.In recent years, with the rapid improvement of urban construction quality, more and more investment is put into urban underground space detection, and high-precision detection of shallow geological structure is required;Mainly including subway track planning, cavity detection under road and drainage pipe maintenance, etc.Small loop TEM system plays an important role in urban underground space geological structure detection due to its small area, high efficiency and large detection depth range.

[0003] Small loop TEM system is usually composed of a transmitter, a multi-channel acquisition device, a transmitting coil and an induction coil.Small loop TEM system generally selects central loop working mode, and Bucking compensation coil is set to reduce the primary induced voltage of receiving coil in order to meet the range of receiving system.The bipolar current excitation of transmitting coil produces a changing primary field, and the eddy current in underground conductor produces a changing secondary field after the transmitting current is turned off;The induction coil acts as a sensor for primary field and secondary field, outputs induced voltage and records it through acquisition device.The electromagnetic response data obtained by acquisition device is processed by data interpretation method to obtain the resistivity distribution of underground medium, so as to obtain the geological structure information of detection area.

[0004] The traditional small-loop time-domain electromagnetic method has low detection accuracy for shallow targets such as urban underground space due to many problems in numerical simulation accuracy, detection instrument and data interpretation method. In the aspect of numerical simulation, the initial field method is used to load the excitation source in the traditional method, and the shallow information is covered. Moreover, the numerical simulation accuracy of small-loop electromagnetic response is low, and the shallow detection effect of different current waveforms cannot be effectively simulated. In the aspect of detection instrument, the output waveform of the traditional transmitting system is a single trapezoidal wave, the electromagnetic wave propagates too fast, the resolution for near-surface targets is low, and there is a detection blind area. The traditional receiving system has problems such as low sampling rate, large noise level and limited acquisition time length, which cannot simultaneously collect the transmitting current and electromagnetic response, and cannot collect full waveform, resulting in low data quality and being not conducive to data interpretation. In the aspect of data interpretation, the traditional data interpretation method mainly interprets the data after the trapezoidal wave is turned off, and it is difficult to interpret the full waveform response, and it is even more difficult to jointly interpret the combined waveform response. The low accuracy of small-loop numerical simulation, low performance of electromagnetic detection instrument system and rigid data interpretation method seriously limit the further development and application of transient electromagnetic method in urban underground space.

[0005] In order to perform high-precision numerical simulation of the detection capability of small-loop, the CN-FDTD method is used to discretize the electromagnetic wave equation, increase the iteration time step and numerical simulation stability. The transmitting current waveform is converted into current density for source-included calculation, realizing numerical simulation of different transmitting current waveforms. The CFS-PML boundary condition is loaded at the same time to further reduce the numerical simulation error, and finally the high-precision numerical simulation of the detection capability of small-loop transmitting coil in urban underground space under different waveform excitation is realized. In addition, the concept of induced electromotive force "cumulative sensitivity" is proposed to quantitatively analyze the detection capability of multiple current waveforms. The resolution of different waveforms to different depth targets is determined by the curve distribution characteristics of the cumulative sensitivity of 90%, so as to select the optimal combination strategy of transmitting parameters and realize high-resolution detection of urban underground space.

[0006] In order to meet the demand of large-scale high-precision detection of urban underground space, the traditional single trapezoidal wave cannot meet the requirements. The half-sine wave has a wide pulse width, and the electromagnetic wave on the falling edge of the half-sine wave is distributed near the surface for a long time, which can effectively obtain the information of shallow targets less than 5m. After the trapezoidal wave transmitting current is turned off, the electromagnetic wave propagates fast, which can effectively identify deeper targets greater than 5m. The trapezoidal-half-sine wave combined transmitting current can improve the detection range and detection accuracy of urban underground space underground structure, and reduce the detection blind area.

[0007] A dual-channel, dual-sampling-rate, low-noise continuous acquisition system is a key component of urban underground space exploration. To completely remove the primary field, the current waveform and electromagnetic response need to be recorded synchronously, thus requiring the acquisition system to achieve simultaneous acquisition across both channels. Furthermore, to acquire the first valid signal as early as possible, the sampling rate of the acquisition system needs to be high enough. However, high sampling rate acquisition results in higher noise levels, so later responses need to be acquired using a low sampling rate. The combination of high and low sampling rates can increase the length of the valid signal. To improve the accuracy of urban underground space data interpretation, full waveform response is required for data interpretation. Therefore, the acquisition system needs to have continuous acquisition capabilities to record the entire waveform of the transmitted current and electromagnetic response.

[0008] A dual-waveform combined transmission system and a dual-channel continuous acquisition and reception system provide hardware support for the dual-waveform combined detection method in urban underground space, and acquire measured current waveforms and full-waveform electromagnetic response information. However, traditional data interpretation methods cannot extract apparent resistivity parameters from the full-waveform response; furthermore, they cannot perform joint data interpretation of the apparent resistivity parameters of trapezoidal waves and half-sine waves. Neural network algorithms have played a significant role in electromagnetic data interpretation in recent years. After training on a sample set, they can extract parameters of the electromagnetic response at each stage. Therefore, a BP neural network algorithm is used to extract apparent resistivity parameters from the full-waveform response, and combined waveform resistivity imaging is performed based on apparent resistivity weighting coefficients to achieve large-scale, high-precision data interpretation of urban underground space.

[0009] Chinese patent CN117492099B discloses a towed time-frequency combined electromagnetic detection system and method for urban underground spaces. This system achieves simultaneous transmission and acquisition of electromagnetic current in both the time and frequency domains, realizing the joint transmission of waveforms in both domains. This effectively solves the problems of traditional electromagnetic detection devices having a single operating mode and limited detection methods. By using two methods or two waveforms in combination for detection, the detection resolution of urban underground spaces can be improved to a certain extent.

[0010] Chinese patent CN113866835B discloses an electromagnetic transmission system and control method for combining three time-domain waveforms. The transmission system comprises a main control circuit, a transmitting bridge, an RLC series resonant circuit, a passive clamping circuit, and a transmitting coil. It can generate trapezoidal waves and triangular waves with different off-times, half-sine waves with different pulse widths, and their combined waveforms, as well as the transmission current. By improving the excitation source, the overall exploration accuracy of the transient electromagnetic method is improved. The transmission of combined waveforms can be achieved by modifying the topology of the transmitting circuit.

[0011] The methods described above, which employ combined time-domain and frequency-domain transmission and combined time-domain polygon transmission to improve the detection accuracy of underground targets, all aim to enhance detection accuracy by improving the transmission source. However, for refined electromagnetic detection of urban underground geological structures, new requirements are placed on small-loop numerical simulation methods, transmission systems, acquisition devices, and data interpretation methods. Improving the transmission system alone has limited effect on improving detection accuracy; existing electromagnetic detection systems are almost entirely unable to achieve large-scale, high-precision detection of urban underground spaces. Furthermore, the sampling rate and noise level of the acquisition device essentially determine the effectiveness of electromagnetic methods for detecting near-surface targets in urban underground spaces. How to output high-quality combined current waveforms, acquire full-waveform electromagnetic responses over long time windows, and employ intelligent algorithms for dual-waveform joint data interpretation to achieve refined detection of urban underground spaces is a pressing technical problem that those skilled in the art must solve. Summary of the Invention

[0012] The technical problem to be solved by this invention is to provide a time-domain dual-waveform combination electromagnetic detection method for urban underground space. The purpose is to achieve refined detection of urban underground space by performing steps such as small loop transmission current waveform detection effect analysis, development of dual-waveform combination transmission system, development of dual-channel continuous acquisition device, and dual-waveform joint data interpretation.

[0013] This invention includes the following steps: a time-domain dual-waveform combination electromagnetic detection method for urban underground space.

[0014] 1) The electromagnetic wave equation is discretized using the CN-FDTD method, and the transmitted current waveform is converted into current density for source-containing calculation. At the same time, CFS-PML boundary conditions are applied to realize high-precision numerical simulation of the ability of small loop transmitting coil to detect urban underground space under different waveform excitation.

[0015] 2) Based on the concept of "cumulative sensitivity" of induced electromotive force, the detection capability of different emission current waveforms is quantitatively analyzed. By observing the distribution curve characteristics with a cumulative sensitivity of 90%, the resolution of different waveforms for targets at different depths is determined. A combination strategy is proposed to measure near-surface targets with the falling edge of the half-sine wave and measure deeper targets after the trapezoidal wave is turned off. The detection range and accuracy of urban underground space are improved by using the trapezoidal-half-sine wave combination.

[0016] 3) The trapezoidal-sine wave combined transmitting circuit topology is designed using H-bridge inverter technology, passive clamping technology, RLC series resonant technology and current isolation technology to realize the combined transmission of half sine wave and trapezoidal wave. The transmission parameters can be adjusted and the two waveforms do not interfere with each other, which significantly improves the excitation capability and vertical detection resolution of urban underground space.

[0017] 4) A dual-channel, dual-sampling-rate, low-noise continuous acquisition system based on a multi-channel variable sampling rate ADC module is designed to achieve synchronous recording of the combined emission current waveform and electromagnetic response; the dual sampling rate acquisition method extends the observation time of the effective signal and improves the detection range; the continuous acquisition method obtains the full waveform response, providing technical support for the subsequent interpretation of the full waveform data;

[0018] 5) The BP neural network algorithm is used to extract the apparent resistivity parameters of the full waveform response of the trapezoidal-semi-sine wave combined emission current. Based on the cumulative sensitivity numerical simulation results, the apparent resistivity weighting coefficients of the two waveforms are determined, and the combined waveform joint resistivity imaging is performed to realize large-scale, high-precision data interpretation of urban underground space.

[0019] In step 2), to analyze the detection capabilities of half-sine waves and trapezoidal waves under small loop conditions, the detection capabilities of the two current waveforms are first quantitatively analyzed based on the concept of "cumulative sensitivity" of induced electromotive force. Specifically, it is expressed as the ratio of the sum of induced electromotive forces Hp at each node from the ground to a certain depth to the sum of induced electromotive forces Ha at all nodes in the calculation area. The mathematical expression for the "cumulative sensitivity" of induced electromotive force is:

[0020]

[0021] In equation (1), U, V, and W represent the total number of mesh nodes in the x, y, and z directions, respectively, w is the number of mesh nodes in the z direction at the current depth, and t n This is the moment of the current iteration number n;

[0022] Based on the quantitative analysis results of numerical simulation of the small loop detection capability of urban underground space using the "cumulative sensitivity" of induced electromotive force, it is clear that using a trapezoidal-semi-sine wave combined emission current as an electromagnetic excitation source can significantly improve the detection range and accuracy of urban underground space.

[0023] In step 3), the trapezoidal-sine wave combined transmitting circuit topology mainly includes an H-bridge inverter circuit, a passive clamping circuit, an absorption circuit, an RLC series resonant circuit, and a current isolation circuit. The passive clamping circuit includes main switching circuits K7 and K8, and two sets of high-voltage clamping circuits K9-K12, TVS-a, and TVS-b based on TVS, used to control the linearity and turn-off time of the trapezoidal wave transmitting current, achieving rapid turn-off of the trapezoidal wave. The absorption circuit consists of a power resistor R0 and switching modules K5 and K6, connected in series with the transmitting coil in the late stage of trapezoidal wave transmitting current turn-off, causing the transmitting coil to operate in a critically damped or over-damped state, suppressing the tail oscillation of the trapezoidal wave transmitting current. The RLC series resonant circuit consists of the bipolar power supply of the H-bridge circuit, resonant capacitor C1, resonant inductor (equivalent to R1 and L2), transmitting coil (equivalent to r and L1), and switching modules K13 and K14. To reduce the complexity of the transmitting circuit and improve stability, the resonant circuit is directly powered by the power supply, generating a bipolar half-sine wave. The amplitude and pulse width of the half-sine wave transmitting current are adjustable. The current isolation circuit consists of two parts: first, four switching modules K1, K2, K3, and K4 of the H-bridge circuit are connected in series with four power diodes D1, D2, D3, and D4, indirectly forming four thyristors. This strictly limits the direction of the half-sine wave transmitting current and suppresses tail oscillations. Second, trapezoidal wave transmitting branches K15 and K16 are set in the transmitting circuit, and two switching modules K13 and K14 are connected in series across the RLC series resonant circuit to achieve current isolation between the trapezoidal wave and the half-sine wave. The two waveforms do not interfere with each other, thus achieving high-quality combined transmission of trapezoidal and half-sine waves, providing a highly stable and high-performance excitation source for urban underground space exploration.

[0024] In step 4), the dual-channel dual-sampling-rate low-noise continuous acquisition system comprises nine parts: a sensor module, a signal conditioning circuit, an analog-to-digital converter (ADC), an analog-to-digital isolation circuit, a main control circuit, a synchronization circuit, a storage circuit, a human-machine interface module, and a power supply module. The sensor module includes a fluxgate current sensor and a receiving coil. The fluxgate current sensor converts the current signal into a voltage signal, and the receiving coil converts the changing magnetic field into an induced voltage. The signal is amplified by a preamplifier before being input to the signal conditioning circuit. The signal conditioning circuit attenuates, filters, and raises the level of the voltage signal output from the sensor, ensuring that the conditioned signal meets the input range requirements of the ADC. The ADC includes a multi-channel variable sampling rate ADC and its peripheral circuitry, simultaneously sampling the two conditioned analog voltage signals. The analog-to-digital isolation circuit mainly includes an analog-to-digital isolation... The isolation unit and its peripheral circuits isolate the analog and digital circuit sections, reducing crosstalk between circuits. The main control circuit generates the system clock and control signals, controlling the system's dual-sampling-rate acquisition, transmit / receive synchronization, data transmission, and storage. The storage circuit includes data transmission and data storage. The digital signal output from the analog-to-digital converter is first transferred to an external SRAM for buffering via DMA, and then the MCU transfers the buffered data to an SD card for storage. Through a ping-pong dual-buffer structure, continuous acquisition and storage of dual-channel data are achieved, obtaining full-waveform electromagnetic response and current waveforms. The human-machine interface circuit mainly includes buttons and a display screen for inputting acquisition commands. The acquisition system achieves high-precision full-waveform synchronous recording of the transmitted current waveform and electromagnetic response, significantly improving the quality of the acquired data and fundamentally enhancing the accuracy of data interpretation.

[0025] In step 5), the BP neural network algorithm for extracting apparent resistivity parameters from the full waveform response of the trapezoidal-semi-sine wave combined emission current mainly includes seven steps: Step 1 is to record the trapezoidal-semi-sine wave combined emission current through a data acquisition device to ensure that the emission current waveform parameters are consistent with the actual detected emission parameters; Step 2 is to prepare the sample set, where the measured trapezoidal wave emission current is convolved with the uniform half-space step response to form sample set 1, and the measured semi-sine wave emission current is convolved with the uniform half-space step response to form sample set 2; the conductivity is set with 2000 values ​​at equal intervals from 0.0001 S / m to 1 S / m, and the time range is set with 1920 values ​​at equal intervals from 10.4 μs to 20 ms to match the electromagnetic response acquisition parameters; finally, an electromagnetic response input sample set of size 2000 × 1920 and a conductivity output sample set of size 2000 × 1920 are obtained; Step 3 is to preprocess the sample set. Since there is a nonlinear relationship between the early electromagnetic response and conductivity, the electromagnetic response and conductivity in the training samples are normalized to between 0.1 and 0.9. Step four involves building the neural network using the "Neural Network Fitting" application in the MATLAB toolbox. The normalized electromagnetic response and conductivity data are imported as the input and output of the neural network, respectively. The number of hidden layers is set to 2, and the number of neurons is set to 8 and 5, respectively. The `newff` function is used to build the network. Step five involves training the neural network using the Levenberg-Marquardt algorithm as the training function. To ensure the generalization ability of the BP neural network, 70% of the total samples are selected as training samples, 15% as validation samples, and 15% as test samples. After training, the mean square error of each time channel is recorded. When the training errors of samples 1 and 2 are both less than 10... -8 Training can only be stopped when the time is right; Step six is ​​to extract the apparent resistivity parameters by substituting the measured full-waveform electromagnetic response into the training set to obtain the apparent resistivity parameters, and at the same time calculate the corresponding apparent depth parameters; Step seven is to determine the apparent resistivity weighting coefficients based on the cumulative sensitivity numerical simulation results. When the apparent depth is less than or equal to 6 meters, the apparent resistivity weighting coefficients for the half-sine wave and the trapezoidal wave are 0.8 and 0.2, respectively; when the apparent depth is greater than 6 meters, the apparent resistivity weighting coefficients for the half-sine wave and the trapezoidal wave are 0.2 and 0.8, respectively; the joint data interpretation of the dual-waveform electromagnetic response is realized through the apparent resistivity weighting coefficients. Attached Figure Description

[0026] Figure 1 This is an overall block diagram of the urban underground space time-domain dual-waveform combination detection system and method of the present invention;

[0027] Figure 2 It is the cumulative sensitivity distribution of small loop half-sine wave and trapezoidal wave in a uniform half-space;

[0028] Figure 3 It is a slice diagram of the numerical simulation of the dual-anomaly model using combined waveforms;

[0029] Figure 4 It is a trapezoidal-sine wave combined transmitting circuit topology;

[0030] Figure 5 It is the measured current waveform of the trapezoidal-semi-sine wave combination;

[0031] Figure 6 This is the overall block diagram of a dual-channel, dual-sampling-rate, low-noise continuous acquisition device;

[0032] Figure 7 These are the test results of the acquisition device for short-circuit noise and standard signals;

[0033] Figure 8 It is a flowchart of dual-waveform joint data interpretation based on BP neural network;

[0034] Figure 9 It consists of the training error of the BP neural network and the apparent resistivity extraction result of the layered model; Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0036] Example

[0037] See Figure 1 This invention provides a time-domain dual-waveform combination electromagnetic detection method for urban underground space, mainly composed of four parts: analysis of the detection effect of small-loop transmitting current waveform, development of a dual-waveform combination transmitting system, development of a dual-channel continuous acquisition device, and joint interpretation of dual-waveform data. First, based on the concept of "cumulative sensitivity" of induced electromotive force, the detection capability of different transmitting current waveforms under small-loop conditions is quantitatively analyzed, providing technical support for the parameter design of the transmitting system. Then, a trapezoidal-sine wave combination transmitting system is designed based on clamping, RLC series resonance, and current isolation techniques. Next, according to the requirement of simultaneously recording the transmitting current waveform and the full waveform of the electromagnetic response, a dual-channel, dual-sampling-rate, low-noise continuous acquisition device is designed. Finally, based on the BP neural network algorithm, the joint interpretation of the dual-waveform combination full waveform response data is completed, realizing large-scale, refined detection of the geological structure of urban underground space.

[0038] See Figure 2 Combination Figure 3As shown, this invention conducted electromagnetic detection numerical simulations on a transmitting coil with a side length of 4 meters. A trapezoidal wave with a turn-off time of 0.2 ms and a half-sine wave with a pulse width of 4 ms were used as excitation sources, respectively. The uniform half-space conductivity was set to 0.01 S / m. The induced electromotive force "cumulative sensitivity" distribution results are as follows: Figure 2 As shown, during the half-sine wave current emission, the curve with a cumulative sensitivity of 90% is distributed for a long time at a depth of 3m, approximately 4ms; furthermore, the curve with a cumulative sensitivity of 90% reaches a depth of 150m in approximately 0.4ms. The trapezoidal wave curve with a cumulative sensitivity of 90% has a very short distribution time near the ground surface; furthermore, the curve with a cumulative sensitivity of 90% reaches a depth of 150m in approximately 0.6ms. The comparison shows that the half-sine wave emission provides higher resolution for near-surface targets at depths less than 3m, while the trapezoidal wave emission provides higher resolution for targets deeper than 3m after the current is turned off. To verify the combined waveform's ability to identify anomalies at different depths, a high-resistivity target was placed at a depth of 1.6m, and a low-resistivity target was placed at a depth of 51.6m. Slices with iteration times less than 10ms were considered trapezoidal wave results, while slices with iteration times greater than 10ms were considered half-sine wave results. Figure 3 As shown, the results of the anomaly model further demonstrate that the falling edge of the half-sine wave has a higher resolution for near-surface targets, and the trapezoidal wave, after being turned off, has a higher resolution for deeper targets.

[0039] See Figure 4 and Figure 5 A trapezoidal-half-sine wave combined transmitting circuit topology was designed based on clamping, RLC series resonance, and current isolation techniques. The RLC resonant circuit is directly powered by the power supply, simplifying the transmitting circuit. Furthermore, by applying current isolation technology, the trapezoidal wave and half-sine wave are designed to prevent interference, significantly improving the quality of the transmitted current waveform. Figure 5 It can be seen that, with the resonant parameters fixed and the clamping voltage changed, the trapezoidal wave amplitude is 36.2A, and the turn-off times are 0.17ms and 0.26ms, respectively; with the clamping voltage fixed and the resonant capacitor changed, the half-sine wave amplitude is 15.6A and 42.7A, respectively, and the pulse width is 5.9ms and 4.6ms, respectively.

[0040] See Figure 6 and Figure 7 The dual-channel, dual-sampling-rate, low-noise continuous acquisition system comprises nine parts: a sensor module, a signal conditioning circuit, an analog-to-digital conversion circuit, an analog-to-digital isolation circuit, a main control circuit, a synchronization circuit, a storage circuit, a human-machine interface module, and a power supply module. The analog-to-digital conversion circuit uses an AK5394 module, enabling dual-channel, dual-sampling-rate acquisition. The buffer circuit employs an IS62WV51216 buffer chip to store two digital signals to an SD card, achieving continuous recording of the entire waveform of the electromagnetic response. Figure 7It can be seen that the baselines of both channels of the acquisition system are near 0, with virtually no baseline drift; the data length is 0.64s, enabling continuous acquisition; the root mean square errors of short-circuit noise are 0.304μV and 0.659μV, respectively. Furthermore, it can perform high-precision acquisition of a standard sine wave signal with an amplitude of 1V and a frequency of 1kHz, and the data consistency between the two channels and the two sampling rates is also very good.

[0041] See Figure 8 and Figure 9 The dual-waveform joint data interpretation process based on BP neural network mainly includes seven steps: acquiring the transmit current of the trapezoidal-sinusoidal combined waveform, obtaining the sample set through convolution operation, sample set preprocessing, building the BP neural network, training the BP neural network, extracting the apparent resistivity parameters, and performing joint resistivity imaging based on the apparent depth weight coefficient; among which, the training errors of both the half-sinusoidal wave and trapezoidal wave sample sets are less than 10. -8 This demonstrates that the training accuracy of this BP neural network is high and can meet the accuracy requirements for extracting apparent resistivity from time-domain electromagnetic response. Furthermore, to verify the effectiveness of the BP neural network apparent resistivity parameter extraction method, semi-analytical solutions of H-type and K-type three-layer geodetic models were used for validation. The two semi-analytical solutions were substituted into the trapezoidal wave sample set network to obtain two sets of apparent resistivity parameters, and then the apparent depth was calculated based on the electromagnetic smoke ring theory. The attenuation curves of the semi-analytical solutions and the apparent resistivity extraction results are shown below. Figure 9 As shown. By Figure 9 It can be seen that the apparent resistivity of the two layered models varies with the apparent depth in a manner that is basically consistent with the model setting, which proves the effectiveness of the BP neural network method in extracting apparent resistivity.

[0042] In urban areas, experimental or exploration sites are selected, and a dual-waveform combined excitation is performed using a transmission system. A dual-channel continuous acquisition device is used to acquire the transmitted current waveform and the full-waveform electromagnetic response information. After subtracting the remaining primary field, the pure response of the dual waveforms is substituted into the neural network training set to extract the apparent resistivity parameter. Finally, the joint data interpretation of the dual waveforms is achieved based on the apparent resistivity weighting coefficient. Specifically, when the apparent depth is less than or equal to 6 meters, the apparent resistivity weighting coefficients for the half-sine wave and the trapezoidal wave are 0.8 and 0.2, respectively; when the apparent depth is greater than 6 meters, the apparent resistivity weighting coefficients for the half-sine wave and the trapezoidal wave are 0.2 and 0.8, respectively. Ultimately, this achieves large-scale, high-precision geological structure detection, which can meet the detection needs of most urban underground spaces and underground structures.

[0043] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A time-domain dual-waveform combination electromagnetic detection method for urban underground space, characterized in that, Includes the following steps: 1) The electromagnetic wave equation is discretized using the CN-FDTD method, and the transmitted current waveform is converted into current density for source-containing calculation. At the same time, CFS-PML boundary conditions are applied to realize high-precision numerical simulation of the ability of small loop transmitting coil to detect urban underground space under different waveform excitation. 2) Based on the concept of "cumulative sensitivity" of induced electromotive force, the detection capability of different emission current waveforms is quantitatively analyzed. By observing the distribution curve characteristics with a cumulative sensitivity of 90%, the resolution of different waveforms for targets at different depths is determined. A combination strategy is proposed to measure near-surface targets with the falling edge of the half-sine wave and measure deeper targets after the trapezoidal wave is turned off. The detection range and accuracy of urban underground space are improved by using the trapezoidal-half-sine wave combination. 3) The trapezoidal-sine wave combined transmitting circuit topology is designed using H-bridge inverter technology, passive clamping technology, RLC series resonant technology and current isolation technology to realize the combined transmission of half sine wave and trapezoidal wave. The transmission parameters can be adjusted and the two waveforms do not interfere with each other, which significantly improves the excitation capability and vertical detection resolution of urban underground space. 4) A dual-channel, dual-sampling-rate, low-noise continuous acquisition system is designed based on a multi-channel variable sampling rate ADC module to achieve synchronous recording of combined emission current waveforms and electromagnetic responses; the dual sampling rate acquisition method extends the observation time of effective signals and improves the detection range. By continuously acquiring data, the full waveform response is obtained, providing technical support for the subsequent interpretation of the full waveform data; 5) The BP neural network algorithm is used to extract the apparent resistivity parameters of the full waveform response of the trapezoidal-semi-sine wave combined emission current. Based on the cumulative sensitivity numerical simulation results, the apparent resistivity weighting coefficients of the two waveforms are determined, and the combined waveform joint resistivity imaging is performed to realize large-scale, high-precision data interpretation of urban underground space.

2. The method for electromagnetic detection of urban underground space using a time-domain dual-waveform combination as described in claim 1, characterized in that: In step 2), to analyze the detection capabilities of half-sine waves and trapezoidal waves under small loop conditions, the detection capabilities of the two current waveforms are first quantitatively analyzed based on the concept of "cumulative sensitivity" of induced electromotive force. Specifically, it is expressed as the ratio of the sum of induced electromotive forces Hp at each node from the ground to a certain depth to the sum of induced electromotive forces Ha at all nodes in the calculation area. The mathematical expression for the "cumulative sensitivity" of induced electromotive force is: (1) In equation (1), U, V, and W represent the total number of mesh nodes in the x, y, and z directions, respectively. This represents the number of mesh nodes in the z-direction at the current depth. This is the moment of the current iteration number n; First, a uniform half-space model was set up to analyze the detection capability of the small-loop transmitting coil. The cumulative sensitivity distribution results of the full waveform of the triangular wave, trapezoidal wave, and half-sine wave transmitting current were obtained. By observing the curve distribution with a cumulative sensitivity of 90%, it can be seen that the falling edge of the half-sine wave has a higher detection resolution for near-surface targets in urban underground space, while the trapezoidal wave has a higher detection resolution for deeper targets in urban underground space after being turned off. In order to further verify the detection capability of the trapezoidal-half-sine wave combined transmitting current for dual-anomaly models at different depths, the identification effect of trapezoidal wave and half-sine wave on anomalies at different depths was observed by slice plot. The results show that the falling edge of the half-sine wave can effectively identify high-resistivity bodies composed of near-surface cavities, while the trapezoidal wave can effectively identify low-resistivity bodies composed of deeper water accumulation areas after being turned off. Based on the quantitative analysis results of numerical simulation of the small loop detection capability of urban underground space using the "cumulative sensitivity" of induced electromotive force, it is clear that using a trapezoidal-semi-sine wave combined emission current as an electromagnetic excitation source can significantly improve the detection range and accuracy of urban underground space.

3. The method for electromagnetic detection of urban underground space using a time-domain dual-waveform combination as described in claim 1, characterized in that: In step 3), the trapezoidal-sine wave combined transmitting circuit topology mainly includes an H-bridge inverter circuit, a passive clamping circuit, an absorption circuit, an RLC series resonant circuit, and a current isolation circuit. The passive clamping circuit includes main switching circuits K7 and K8, and two sets of high-voltage clamping circuits K9~K12, TVS-a, and TVS-b based on TVS, used to control the linearity and turn-off time of the trapezoidal wave transmitting current, achieving rapid turn-off of the trapezoidal wave. The absorption circuit consists of a power resistor R0 and switching modules K5 and K6, which control the linearity and turn-off time of the trapezoidal wave transmitting current. The late-stage turn-off circuit is connected in series with the transmitting coil, enabling the transmitting coil to operate in a critically damped or overdamped state, suppressing the tail oscillation of the trapezoidal wave transmitting current. The RLC series resonant circuit consists of a bipolar power supply for the H-bridge, a resonant capacitor C1, a resonant inductor (equivalent to R1 and L2), a transmitting coil (equivalent to r and L1), and switching modules K13 and K14. To reduce the complexity of the transmitting circuit and improve stability, the resonant circuit is directly powered by the power supply, generating a bipolar half-sine wave, and the amplitude and pulse width parameters of the half-sine wave transmitting current can be adjusted. The current isolation circuit consists of two parts. First, the four switching modules K1, K2, K3, and K4 of the H-bridge circuit are connected in series with four power diodes D1, D2, D3, and D4 to indirectly form four thyristors, which can strictly limit the direction of the half-sine wave transmission current and suppress the tail oscillation of the half-sine wave transmission current. Second, trapezoidal wave transmission branches K15 and K16 are set in the transmission circuit, and two switching modules K13 and K14 are connected in series across the RLC series resonant circuit to achieve current isolation between the trapezoidal wave and the half-sine wave. The two waveforms do not interfere with each other, thereby realizing high-quality combined transmission of trapezoidal and half-sine waves, providing a highly stable and high-performance excitation source for urban underground space exploration. In step 4), the dual-channel dual-sampling-rate low-noise continuous acquisition system comprises nine parts: a sensor module, a signal conditioning circuit, an analog-to-digital conversion circuit, an analog-to-digital isolation circuit, a main control circuit, a synchronization circuit, a storage circuit, a human-machine interaction module, and a power supply module. The sensor module includes a fluxgate current sensor and a receiving coil. The fluxgate current sensor converts the current signal into a voltage signal, and the receiving coil converts the changing magnetic field into an induced voltage. The signal is amplified by a preamplifier and then input to the signal conditioning circuit. The signal conditioning circuit is responsible for attenuating, filtering, and level-up the voltage signal output by the sensor, so that the conditioned signal can meet the input range requirements of the ADC. The analog-to-digital conversion circuit includes a multi-channel variable sampling rate ADC and its peripheral circuits, which simultaneously samples the two conditioned analog voltage signals. Analog-to-digital isolation circuits mainly include analog-to-digital isolators and their peripheral circuits, which isolate the analog circuit part and the digital circuit part to reduce crosstalk between circuits; The main control circuit generates the system clock and control signals, controlling the system's dual-sampling-rate acquisition, transmit / receive synchronization, data transmission, and storage. The storage circuit includes data transmission and data storage. The digital signal output from the analog-to-digital converter is first transferred to an external SRAM for buffering via DMA. Then, the MCU transfers the buffered data to an SD card for storage. Through a ping-pong dual-buffer structure, continuous acquisition and storage of dual-channel data are achieved, obtaining full-waveform electromagnetic response and current waveforms. The human-machine interface circuit mainly includes buttons and a display screen for inputting acquisition commands. The acquisition system achieves high-precision full-waveform synchronous recording of the transmitted current waveform and electromagnetic response, significantly improving the quality of the acquired data and fundamentally enhancing the accuracy of data interpretation. The trapezoidal-semi-sine wave combined transmission system and the dual-channel low-noise continuous acquisition system provide hardware support for the refined exploration of urban underground space.

4. The method for electromagnetic detection of urban underground space using a time-domain dual-waveform combination as described in claim 1, characterized in that: In step 5), the BP neural network algorithm for extracting apparent resistivity parameters from the full waveform response of the trapezoidal-sinusoidal combined emission current mainly includes seven steps. Step one involves recording the trapezoidal-sinusoidal combined emission current using a data acquisition device to ensure that the emission current waveform parameters are consistent with the actual detected emission parameters. Step two involves sample set preparation: the measured trapezoidal wave emission current is convolved with the uniform half-space step response to form sample set 1, and the measured half-sinusoidal wave emission current is convolved with the uniform half-space step response to form sample set 2. The conductivity is set at 2000 equal intervals from 0.0001 S / m to 1 S / m, and the time range is from 10.4 μs to 20 μs to match the electromagnetic response acquisition parameters. The values ​​are set to 1920 at equal intervals (ms). Finally, an electromagnetic response input sample set of size 2000×1920 and an conductivity output sample set of size 2000×1920 are obtained. Step three involves preprocessing the sample sets. Due to the nonlinear relationship between the early electromagnetic response and conductivity, both the electromagnetic response and conductivity in the training samples are normalized to between 0.1 and 0.

9. Step four involves building the neural network. The "Neural Network Fitting" application in the MATLAB toolbox is used to build the neural network. The normalized electromagnetic response and conductivity data are imported as the input and output of the neural network, respectively. The number of hidden layers in the neural network is set to 2, and the number of neurons is set to 8 and 5, respectively. The newff function is used to build the network. Step five involves training the neural network. The Levenberg-Marquardt algorithm is used as the training function. To ensure the generalization ability of the BP neural network, 70% of the total samples are selected as training samples, 15% as validation samples, and 15% as test samples. After training, record the mean square error for each time channel. When the training errors of both sample sets 1 and 2 are less than 10... -8 Training can only be stopped when the time is right; Step six is ​​to extract the apparent resistivity parameters by substituting the measured full-waveform electromagnetic response into the training set to obtain the apparent resistivity parameters, and at the same time calculate the corresponding apparent depth parameters; Step seven is to determine the apparent resistivity weighting coefficients based on the cumulative sensitivity numerical simulation results. When the apparent depth is less than or equal to 6 meters, the apparent resistivity weighting coefficients for the half-sine wave and the trapezoidal wave are 0.8 and 0.2, respectively; when the apparent depth is greater than 6 meters, the apparent resistivity weighting coefficients for the half-sine wave and the trapezoidal wave are 0.2 and 0.8, respectively; the joint data interpretation of the dual-waveform electromagnetic response is realized through the apparent resistivity weighting coefficients.

Images