In-tunnel safety monitoring method based on multiple sensors

Abstract

The invention discloses an in-tunnel safety monitoring method based on multiple sensors. The method comprises the steps of signal acquisition, data set establishment, tunnel safety monitoring model establishment and tunnel safety monitoring. The invention belongs to the technical field of tunnel safety monitoring, and particularly relates to an in-tunnel safety monitoring method based on multiple sensors, and the method comprises the steps: extracting signal information through singular spectrum decomposition and residual signal processing; hilbert transformation and phase sequence calculation are used, and phase information of an original signal is reserved; judgment is carried out according to the complexity of the residual signals, and the adaptability and efficiency of signal decomposition are improved; adjustable parameters are introduced when an activation function is designed, and in a negative number range, an exponential decay function is adopted, so that negative number input is effectively mapped to a relatively small output range, and the gradient explosion problem is avoided; designing an initial movement strategy by balancing random search and fine search, and randomly selecting a part of individuals of a population to move randomly; and the search convergence speed and quality are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel safety monitoring, and specifically refers to a tunnel safety monitoring method based on multiple sensors. Background Art

[0002] The tunnel safety monitoring method is a technical method that uses sensors, data processing technologies, and model establishment and optimization to monitor and evaluate the tunnel structure, environment, and operating status in real time. Its main purpose is to identify potential safety hazards in the tunnel, take preventive and control measures in a timely manner, and ensure the safe operation of the tunnel. However, in general, the tunnel safety monitoring method has problems such as excessive noise content in the original signal, over-decomposition during signal decomposition, and poor ability to retain signal information; in general, the tunnel safety monitoring method has problems such as unstable model learning process due to improper design of the model activation function, slow model convergence speed due to improper setting of model parameters, and low accuracy when searching for parameters. Summary of the Invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a tunnel safety monitoring method based on multiple sensors. Aiming at the problems of excessive noise content in the original signal, over-decomposition during signal decomposition, and poor ability to retain signal information in the general tunnel safety monitoring method, this solution extracts signal information through singular spectrum decomposition and residual signal processing; uses Hilbert transform and phase sequence calculation to retain the phase information of the original signal; makes a judgment according to the complexity of the residual signal, effectively avoiding over-decomposition of unnecessary signals, and improving the adaptability and efficiency of signal decomposition; aiming at the problems of unstable model learning process due to improper design of the model activation function, slow model convergence speed due to improper setting of model parameters, and low accuracy when searching for parameters in the general tunnel safety monitoring method, this solution introduces adjustable parameters when designing the activation function, and adopts an exponential decay function in the negative range, making the model learning process more flexible, effectively mapping negative inputs to a smaller output range, and avoiding the problem of gradient explosion; designs the initial movement strategy by weighing random search and refined search, and randomly selects a part of the individuals in the population for random movement; improves the search convergence speed and quality; and further improves the model accuracy and establishment speed.

[0004] The technical solution adopted by the present invention is as follows: The tunnel safety monitoring method based on multiple sensors provided by the present invention includes the following steps: Step S1: Signal acquisition; Step S2: Establish a data set; Step S3: Establish a tunnel safety monitoring model; Step S4: Tunnel safety monitoring.

[0005] Further, in step S1, the signal acquisition is based on sensors to collect historical signal data in the tunnel; the historical signal data in the tunnel includes environmental parameters, structural parameters, gas concentration, traffic flow, sound signals, and tunnel safety status; the tunnel safety status is used as a data label.

[0006] Further, in step S2, the establishment of the data set specifically includes the following steps: Step S21: Pre-decomposition; decompose the signal to be decomposed based on singular spectrum decomposition, set the number of components to 1, and use the only single singular spectrum component obtained by decomposition as the result; after k iterations, the component is obtained; it is expressed as follows: ; In the formula, SSC k is the singular spectrum component obtained after k iterations; y is the signal to be decomposed; SSD(·) is the singular spectrum decomposition process; Step S22: Obtain the residual signal. Based on the residual signal, calculate the error between the residual signal and the original signal in the k-th iteration. There is a preset error threshold. When the error is lower than the error threshold, the decomposition is completed, and go to step S24; otherwise, go to step S23; the formula used is as follows: ; In the formula, res is the residual signal, and SSC i1 is the singular spectrum component at the i1-th iteration; Step S23: Evaluate complexity; there is a preset complexity threshold, which is judged after each signal decomposition. When the complexity of the singular spectrum component is higher than the complexity threshold, the residual signal is judged as a useless signal; otherwise, the residual signal is regarded as the original signal for the next decomposition; specifically including: Step S231: Hilbert transform, the formula used is as follows: ; ; In the formula, Z(i2) is the transformed complex sequence; Y(i2) is the signal to be decomposed; H[·] is the Hilbert transform; j is the imaginary unit; A(i2) is the amplitude part; g(i2) is the phase angle; is the value of the signal to be processed at the time delay τ; Step S232: Calculate the phase sequence, the formula used is as follows: ; In the formula, θ(·) is the element of the phase sequence, and i2 is the index of the element of the phase sequence; Step S233: Reconstruct the phase space. Based on permutation entropy, use the phase information of each signal to reconstruct the phase space; the formula used is as follows: ; Wherein, S(j1) is the reconstructed component in the reconstructed phase space, and j1 is the starting position index; m is the reconstruction parameter, representing the number of phase angles included in each reconstructed component; Step S234: Arrange each reconstructed component in ascending order of numerical value, extract the position index of each element in the original reconstructed sequence in ascending order for S(j), and record the position index as l, thereby obtaining the corresponding position sequence; and calculate the occurrence probability of the same position sequence; the formula used is as follows: ; Wherein, P i2 is the occurrence probability of the i2th same position sequence; Num i2 is the number of the same position sequences; N is the total number of the reconstructed sequences; Step S235: Evaluate the signal complexity, and the formula used is as follows: ; Wherein, PG is the evaluation value of the signal complexity; p is the normalization parameter; q is the expansion parameter, which is used to improve the sensitivity to extreme probabilities; is the weight parameter, which is used to adjust the probability weighting; sign(·) is the sign function; Step S24: Divide the data set; obtain the data set based on the decomposed signal and divide it into a training set and a test set.

[0007] Furthermore, in step S3, the establishment of the tunnel safety monitoring model is to establish a BP neural network based on the data set divided in step S2 and optimize the model parameters; specifically, it includes the following steps: Step S31: Initialize, initialize the connection weights and biases of the neurons; Step S32: Forward propagation, first calculate the input of the neurons in the hidden layer, and then perform a non-linear transformation on the input through the activation function to obtain the output of the neurons in the hidden layer; it is expressed as follows: ; ; Wherein, u i is the input of the i-th neuron in the hidden layer; ω mi is the connection weight from the m-th neuron in the input layer to the i-th neuron in the hidden layer; b is the bias of the hidden layer; v i is the output of the i-th neuron in the hidden layer; f(·) is the sigmoid activation function; M is the total number of neurons; x m is the input of the m-th neuron in the input layer; Step S33: Calculate the output; the formula used is as follows: ; ; ; wherein, u j is the input of the j-th neuron; ω ij is the connection weight from the i-th neuron in the hidden layer to the j-th neuron in the output layer; is the threshold of the i-th neuron in the output layer; g(·) is the activation function; x is the input of the activation function; α and are the activation function parameters that control the shape of the function; Step S34: Backpropagation, update the weights and thresholds reversely according to the error backpropagation; when the model converges or reaches the maximum number of training times, the training of the tunnel safety monitoring model is completed; Step S35: Model parameter optimization, optimize the initial parameters of the neural network and the activation function parameters; specifically including: Step S351: Optimization initialization, establish a search space based on the initial parameters of the neural network and the activation function parameters; randomly initialize the optimization population, and use the prediction accuracy rate of the tunnel safety monitoring model trained based on the position of the optimization individual for the test set as the fitness value of the optimization individual; Step S352: Design the initial movement strategy, select the top 10% of the individuals with the fitness value as the preferred individuals; balance the random search and the refined search, and the formula used is as follows: ; wherein, and are the positions of the i-th optimization individual in the d-th dimension at the (t + 1)-th iteration and the t-th iteration respectively; r1 is a random number in the range of 0 to 1; f i is the fitness value of the optimization individual; f best is the maximum fitness value of the population; is the position of the i-th preferred individual in the d-th dimension at the t-th iteration; BI is the total number of preferred individuals; is the historical optimal position of the optimization individual in the d-th dimension at the t-th iteration; β is a random parameter; γ is the movement decision threshold; Step S353: Random movement, randomly select 10% of the individuals in the population for random movement, and the formula used is as follows: ; wherein, is the position of the optimization individual after random movement; r2 is a random number in the range of 0 to 1, independent of r1; far(·) is the position of the individual farthest from the target individual; Step S354: Judgment. A fitness threshold is preset. When there is an optimized individual whose fitness is higher than the fitness threshold, the establishment of the tunnel safety monitoring model is completed; if the maximum number of iterations is reached, return to step S351; otherwise, continue the iterative search.

[0008] Further, in step S4, the tunnel safety monitoring is based on the established tunnel safety monitoring model, and real-time tunnel signal data is collected by sensors. The real-time tunnel signal data includes environmental parameters, structural parameters, gas concentration, traffic flow, and sound signals; after signal decomposition, it is input into the tunnel safety monitoring model; and tunnel safety monitoring is realized based on the output of the tunnel safety monitoring model.

[0009] The beneficial effects achieved by the present invention using the above solution are as follows:

[0010] (1) Aiming at the problems that the general tunnel safety monitoring method has a too large noise content in the original signal, over-decomposition exists in the signal decomposition process, and the ability to retain signal information is poor, this solution extracts signal information through singular spectrum decomposition and residual signal processing; uses Hilbert transform and the calculation of phase sequences to retain the phase information of the original signal; makes a judgment according to the complexity of the residual signal, effectively avoiding over-decomposition of unnecessary signals, and improving the adaptability and efficiency of signal decomposition.

[0011] (2) Aiming at the problems that the general tunnel safety monitoring method has an unstable model learning process due to improper design of the model activation function, a slow model convergence speed and low accuracy when searching for parameters due to improper model parameter settings, this solution introduces adjustable parameters when designing the activation function. In the negative range, an exponential decay function is used, making the model learning process more flexible, effectively mapping negative inputs to a smaller output range, and avoiding the problem of gradient explosion; designs the initial movement strategy by weighing random search and refined search, and randomly selects a part of the individuals in the population for random movement; improves the search convergence speed and quality; and further improves the model accuracy and establishment speed. Description of the Drawings

[0012] Figure 1 It is a schematic flow chart of the tunnel safety monitoring method based on multiple sensors provided by the present invention; Figure 2 It is a schematic flow chart of step S2; Figure 3 It is a schematic flow chart of step S3; The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention. Detailed Embodiments

[0013] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0014] In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation to the present invention.

[0015] Embodiment 1. Refer to Figure 1 , the tunnel safety monitoring method based on multiple sensors provided by the present invention includes the following steps: Step S1: Signal acquisition. Based on the sensors, historical signal data in the tunnel is collected. Step S2: Establish a data set. Signal information is extracted through singular spectrum decomposition and residual signal processing, and the phase information of the original signal is retained through Hilbert transform and calculation of the phase sequence, thereby completing the establishment of the data set. Step S3: Establish a tunnel safety monitoring model. By introducing adjustable parameters when designing the activation function, in the negative range, an exponential decay function is adopted; an initial movement strategy is designed by weighing random search and refined search, and a part of the individuals in the population are randomly selected for random movement to optimize the model parameters; thereby establishing a tunnel safety monitoring model. Step S4: Tunnel safety monitoring.

[0016] Embodiment 2. Refer to Figure 1 , based on the above embodiment, in step S1, the historical signal data in the tunnel includes environmental parameters, structural parameters, gas concentration, traffic flow, sound signals, and tunnel safety status; the tunnel safety status is used as the data label.

[0017] Embodiment 3. Refer to Figure 1 and Figure 2 , based on the above embodiment, in step S2, the establishment of the data set specifically includes the following steps: Step S21: Pre-decomposition. Based on singular spectrum decomposition, the signal to be decomposed is decomposed, the number of components is set to 1, and the only single singular spectrum component obtained by decomposition is used as the result; after k iterations, the component is obtained; it is expressed as follows: ; In the formula, SSC k is the singular spectrum component obtained after k iterations; y is the signal to be decomposed; SSD(·) is the singular spectrum decomposition process; Step S22: Obtain the residual signal. Based on the residual signal, calculate the error between the residual signal and the original signal in the k-th iteration. There is a preset error threshold. When the error is lower than the error threshold, the decomposition is completed, and go to step S24; otherwise, go to step S23; The formula used is as follows: ; In the formula, res is the residual signal, and SSC i1 is the singular spectrum component at the i1-th iteration; Step S23: Evaluate the complexity; There is a preset complexity threshold, which is determined after each signal decomposition. When the complexity of the singular spectrum component is higher than the complexity threshold, the residual signal is determined to be a useless signal; otherwise, the residual signal is regarded as the original signal for the next decomposition; Specifically include: Step S231: Hilbert transform, the formula used is as follows: ; ; In the formula, Z(i2) is the transformed complex sequence; Y(i2) is the signal to be decomposed; H[·] is the Hilbert transform; j is the imaginary unit; A(i2) is the amplitude part; g(i2) is the phase angle; is the value of the signal to be processed at the time delay τ; Step S232: Calculate the phase sequence, the formula used is as follows: ; In the formula, θ(·) is the phase sequence element, and i2 is the phase sequence element index; Step S233: Reconstruct the phase space. Based on the permutation entropy, use the phase information of each signal to reconstruct the phase space; The formula used is as follows: ; In the formula, S(j1) is the reconstruction component in the reconstructed phase space, and j1 is the starting position index; m is the reconstruction parameter, indicating the number of phase angles included in each reconstruction component; Step S234: Arrange each reconstruction component in ascending order of numerical value. S(j) extracts the position index of each element in the original reconstruction sequence in ascending order, and record the position index as l, so as to obtain the corresponding position sequence; And calculate the occurrence probability of the same position sequence; The formula used is as follows: ; In the formula, P i2 is the occurrence probability of the i2-th same position sequence; Numi2 is the number of identical position sequences; N is the total number of reconstructed sequences; Step S235: Evaluate the signal complexity using the following formula: ; In the formula, PG is the evaluation value of signal complexity; q is the expansion parameter used to increase the sensitivity to extreme probabilities; is the weight parameter used to adjust probability weighting; sign(·) is the sign function; Step S24: Divide the data set; obtain the data set based on the decomposed signal and divide it into a training set and a test set.

[0018] By performing the above operations, for the problems of excessive noise content in the original signal, over-decomposition during signal decomposition, and poor ability to retain signal information in the general tunnel safety monitoring method, this solution extracts signal information through singular spectrum decomposition and residual signal processing; uses Hilbert transform and calculation of phase sequences to retain the phase information of the original signal; makes a determination based on the complexity of the residual signal, effectively avoiding over-decomposition of unnecessary signals and improving the adaptability and efficiency of signal decomposition.

[0019] Example 4, refer to Figure 1 and Figure 3 , based on the above example, in step S3, establish a tunnel safety monitoring model by establishing a BP neural network based on the data set divided in step S2 and optimize the model parameters; specifically include the following steps: Step S31: Initialize, initialize the connection weights and biases of the neurons; Step S32: Forward propagation, first calculate the input of the neurons in the hidden layer, and then perform a non-linear transformation on the input through the activation function to obtain the output of the neurons in the hidden layer; expressed as follows: ; ; In the formula, u i is the input of the i-th neuron in the hidden layer; ω mi is the connection weight from the m-th neuron in the input layer to the i-th neuron in the hidden layer; b is the bias of the hidden layer; v i is the output of the i-th neuron in the hidden layer; f(·) is the sigmoid activation function; M is the total number of neurons; x m is the input of the m-th neuron in the input layer; Step S33: Calculate the output; use the following formula: ; ; ; In the formula, u j is the input of the j-th neuron; ω ij is the connection weight from the i-th neuron in the hidden layer to the j-th neuron in the output layer; is the threshold of the i-th neuron in the output layer; g(·) is the activation function; x is the input of the activation function; α and are the activation function parameters that control the shape of the function; Step S34: Backpropagation. Update the weights and thresholds inversely according to the error backpropagation. When the model converges or reaches the maximum number of training times, the training of the tunnel safety monitoring model is completed; Step S35: Model parameter optimization. Optimize the initial parameters of the neural network and the activation function parameters. Specifically, it includes: Step S351: Optimization initialization. Establish a search space based on the initial parameters of the neural network and the activation function parameters. Randomly initialize the optimization population, and use the prediction accuracy rate of the tunnel safety monitoring model trained based on the position of the optimization individual on the test set as the fitness value of the optimization individual; Step S352: Design the initial movement strategy. Select the top 10% of the individuals with the fitness value as the preferred individuals. Weigh the random search and the refined search. The formula used is as follows: ; In the formula, and are the positions of the i-th optimization individual in the d-th dimension at the (t + 1)-th iteration and the t-th iteration respectively; r1 is a random number in the range of 0 to 1; f i is the fitness value of the optimization individual; f best is the maximum fitness value of the population; is the position of the i-th preferred individual in the d-th dimension at the t-th iteration; BI is the total number of preferred individuals; is the historical optimal position of the optimization individual in the d-th dimension at the t-th iteration; β is a random parameter; γ is the movement decision threshold; Step S353: Random movement. Randomly select 10% of the individuals in the population for random movement. The formula used is as follows: ; In the formula, is the position of the optimization individual after random movement; r2 is a random number in the range of 0 to 1, independent of r1; far(·) is the position of the individual farthest from the target individual; Step S354: Judgment. There is a preset fitness threshold. When there is an optimization individual with a fitness threshold higher than the fitness threshold, the establishment of the tunnel safety monitoring model is completed; if the maximum number of iterations is reached, return to Step S351; otherwise, continue the iterative search.

[0020] By performing the above operations, aiming at the problems existing in the general tunnel safety monitoring method, such as the improper design of the model activation function leading to an unstable model learning process, the improper setting of model parameters resulting in a slow model convergence speed and low accuracy when searching for parameters, this solution introduces adjustable parameters when designing the activation function. In the negative range, an exponential decay function is adopted, making the model learning process more flexible, effectively mapping negative inputs to a smaller output range, and avoiding the problem of gradient explosion. By weighing random search and refined search, an initial movement strategy is designed, and a part of the individuals in the population are randomly selected for random movement, improving the search convergence speed and quality, and thus improving the model accuracy and establishment speed.

[0021] Example 5, refer to Figure 1 , based on the above example, in step S4, for the established tunnel safety monitoring model, real-time tunnel signal data is collected based on sensors. The real-time tunnel signal data includes environmental parameters, structural parameters, gas concentration, traffic flow, and sound signals. After signal decomposition, it is input into the tunnel safety monitoring model, and tunnel safety monitoring is realized based on the output of the tunnel safety monitoring model.

[0022] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.

[0023] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention.

[0024] The above describes the present invention and its implementation manners. This description is not restrictive. What is shown in the drawings is only one of the implementation manners of the present invention, and the actual structure is not limited thereto. Generally speaking, if those of ordinary skill in the art are inspired by it and design similar structural manners and embodiments without creative work without departing from the purpose of the present invention, they should all fall within the protection scope of the present invention.

Claims

1. A safety monitoring method in a tunnel based on multiple sensors, characterized in that: The method includes the following steps: Step S1: Signal acquisition; Step S2: Establish a data set. Extract signal information through singular spectrum decomposition and residual signal processing, and use Hilbert transform and calculation of phase sequences to retain the phase information of the original signal, thereby establishing a data set; Step S3: Establish a tunnel safety monitoring model. By introducing adjustable parameters when designing the activation function, in the negative range, use an exponential decay function; design an initial movement strategy by weighing random search and refined search, and randomly select a part of the individuals in the population for random movement to optimize the model parameters; thereby establish a tunnel safety monitoring model; Step S4: Tunnel safety monitoring.

2. The method for in-tunnel safety monitoring based on multiple sensors according to claim 1, wherein: In step S2, the establishment of the data set specifically includes the following steps: Step S21: Pre-decomposition; Decompose the signal to be decomposed based on singular spectrum decomposition, set the number of components to 1, and use the only single singular spectrum component obtained by decomposition as the result; after k iterations, obtain the component; expressed as follows: ； where SSC k is the singular spectrum component obtained after k iterations; y is the signal to be decomposed; SSD(·) is the singular spectrum decomposition process; Step S22: Obtain the residual signal. Based on the residual signal, calculate the error between the residual signal and the original signal in the k-th iteration. There is a preset error threshold. When the error is lower than the error threshold, the decomposition is completed, and go to step S24; otherwise, go to step S23; the formula used is as follows: ； where res is the residual signal, and SSC i1 is the singular spectrum component at the i1-th iteration; Step S23: Evaluate complexity; Step S24: Divide the data set; Obtain the data set based on the decomposed signal and divide it into a training set and a test set.

3. The method for in-tunnel safety monitoring based on multiple sensors according to claim 2, characterized in that: In step S23, the evaluation of complexity is that there is a preset complexity threshold, which is determined after each signal decomposition. When the complexity of the singular spectrum component is higher than the complexity threshold, the residual signal is determined to be a useless signal; otherwise, the residual signal is regarded as the original signal for the next decomposition; specifically includes: Step S231: Hilbert transform, the formula used is as follows: ； ； Where, Z(i2) is the transformed complex sequence; Y(i2) is the signal to be decomposed; H[·] is the Hilbert transform; j is the imaginary unit; A(i2) is the amplitude part; g(i2) is the phase angle; is the value of the signal to be processed at the time delay τ; Step S232: Calculate the phase sequence, the formula used is as follows: ； where θ(·) is the phase sequence element, and i2 is the phase sequence element index; Step S233: Reconstruct the phase space. Based on permutation entropy, use the phase information of each signal to reconstruct the phase space; the formula used is as follows: ； where S(j1) is the reconstructed component in the reconstructed phase space, j1 is the starting position index; m is the reconstruction parameter, indicating the number of phase angles included in each reconstructed component; Step S234: Arrange each reconstructed component in ascending order of value. S(j) extracts the position index of each element in the original reconstructed sequence in ascending order, and record the position index as l, thereby obtaining the corresponding position sequence; and calculate the occurrence probability of the same position sequence; the formula used is as follows: ； where P i2 is the occurrence probability of the i2-th same position sequence; Num i2 is the number of the same position sequences; N is the number of the total reconstructed sequences; Step S235: Evaluate signal complexity, the formula used is as follows: ； where PG is the evaluation value of signal complexity; q is the expansion parameter used to improve the sensitivity to extreme probabilities; is the weight parameter used to adjust probability weighting; sign(·) is the sign function.

4. The safety monitoring method in a tunnel based on multiple sensors according to claim 1, characterized in that: In step S3, the establishment of the tunnel safety monitoring model is to establish a BP neural network based on the data set divided in step S2 and optimize the model parameters; specifically includes the following steps: Step S31: Initialization, initialize the connection weights and biases of the neurons; Step S32: Forward propagation, first calculate the input of the neurons in the hidden layer, and then perform a non-linear transformation on the input through the activation function to obtain the output of the neurons in the hidden layer; expressed as follows: ； ； where u i is the input of the i-th hidden layer neuron; ω mi is the connection weight from the m-th neuron in the input layer to the i-th neuron in the hidden layer; b is the bias of the hidden layer; v i is the output of the i-th neuron in the hidden layer; f(·) is the sigmoid activation function; M is the total number of neurons; x m is the input of the m-th neuron in the input layer; Step S33: Calculate the output; the formula used is as follows: ； ； ； where u j is the input of the j-th neuron; ω ij is the connection weight from the i-th neuron in the hidden layer to the j-th neuron in the output layer; is the threshold of the i-th neuron in the output layer; g(·) is the activation function; x is the input of the activation function; α and are the activation function parameters that control the shape of the function; Step S34: Backpropagation. Update the weights and thresholds reversely according to the error. When the model converges or reaches the maximum number of training times, the training of the tunnel safety monitoring model is completed. Step S35: Optimize the model parameters.

5. The safety monitoring method in a tunnel based on multiple sensors according to claim 4, characterized in that: The model parameter optimization described in step S35 is to optimize the initial parameters of the neural network and the parameters of the activation function. Specifically, it includes: Step S351: Optimization initialization. Establish a search space based on the initial parameters of the neural network and the parameters of the activation function. Randomly initialize the optimization population, and use the prediction accuracy rate of the tunnel safety monitoring model trained based on the positions of the optimized individuals for the test set as the fitness value of the optimized individuals. Step S352: Design the initial movement strategy. Select the top 10% of the individuals with the fitness value as the preferred individuals. Weigh the random search and the refined search. The formula used is as follows: ； In the formula, and are the positions of the d - dimension of the i - th optimized individual at the (t + 1)-th iteration and the t - th iteration respectively; r1 is a random number in the range from 0 to 1; f i is the fitness value of the optimized individual; f best is the maximum fitness value of the population; is the position of the d - dimension of the bi - th preferred individual at the t - th iteration; BI is the total number of preferred individuals; is the historical optimal position of the optimized individual of the d - dimension at the t - th iteration; β is a random parameter; γ is the movement decision threshold; Step S353: Random movement. Randomly select 10% of the individuals in the population for random movement. The formula used is as follows: ； In the formula, is the optimized individual position after random movement; r2 is a random number in the range from 0 to 1 and is independent of r1; far(·) is the position of the individual that is farthest from the target individual; Step S354: Judgment. There is a preset fitness threshold. When there is an optimized individual with a fitness value higher than the fitness threshold, the establishment of the tunnel safety monitoring model is completed. If the maximum number of iterations is reached, return to step S351. Otherwise, continue the iterative search.

6. The safety monitoring method in a tunnel based on multiple sensors according to claim 1, characterized in that: In step S4, the tunnel safety monitoring is for the established tunnel safety monitoring model, based on the sensor to collect the real-time signal data in the tunnel. The real-time signal data in the tunnel includes environmental parameters, structural parameters, gas concentration, traffic flow, and sound signals. After signal decomposition, it is input into the tunnel safety monitoring model. Based on the output of the tunnel safety monitoring model, realize the safety monitoring in the tunnel.

7. The safety monitoring method in a tunnel based on multiple sensors according to claim 1, characterized in that: In step S1, the signal acquisition is to collect the historical signal data in the tunnel based on the sensor. The historical signal data in the tunnel includes environmental parameters, structural parameters, gas concentration, traffic flow, sound signals, and the tunnel safety state. Use the tunnel safety state as the data label.

Images