Tunnel construction dust treatment control method, system, device and medium

Abstract

The application discloses a tunnel construction dust treatment control method, system, device and medium, and relates to the technical field of tunnel construction dust treatment. Multiple values are obtained in the preset value range of the spraying parameter, multiple spraying parameter data are obtained, for each spraying parameter data, the spraying parameter data and the current tunnel environment data are taken as inputs, a trained prediction model is used to obtain the prediction value of the treated tunnel dust concentration corresponding to the spraying parameter data, the spraying parameter data corresponding to the minimum value of the prediction value of the treated tunnel dust concentration is selected as the target parameter data, and dust treatment is performed based on the target parameter data, so that the spraying parameter data that can make the treated tunnel dust concentration minimum can be selected according to the tunnel environment data to perform dust treatment, the spraying parameter can be automatically and dynamically adjusted according to the actual construction site, the dust reduction effect is improved, and the dust reduction cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of dust control technology in tunnel construction, and in particular to a method, system, equipment and medium for controlling dust in tunnel construction based on machine learning. Background Technology

[0002] Due to the narrow and enclosed nature of tunnels, dust problems are particularly prominent during tunnel construction. Dust inside tunnels can cause wear and tear on construction machinery, reduce the accuracy of instruments, and shorten their lifespan. Furthermore, workers exposed to high concentrations of dust for extended periods are susceptible to occupational dust-related diseases. Therefore, it is crucial to implement reliable dust control technologies in a timely manner during tunnel construction.

[0003] Based on the principles of dust control technology and the different processes of dust diffusion within tunnels, dust removal is categorized into techniques such as "reduction, lowering, emission, removal, and obstruction." Conventional ventilation methods are often ineffective at removing fine dust, while personal protective equipment (PPE) is impractical for strenuous physical labor within tunnels and can easily lead to safety accidents. Research indicates that spray dust suppression technology is the most economical and effective means of dust control. However, in construction sites, spray parameters are fixed and cannot be automatically adjusted according to the actual construction situation, resulting in poor dust suppression effects, water waste, and high costs. Summary of the Invention

[0004] The purpose of this invention is to provide a method, system, equipment, and medium for controlling dust during tunnel construction, which can automatically and dynamically adjust spray parameters according to the actual construction site, thereby improving dust suppression effect and reducing dust suppression costs.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] A method for controlling dust during tunnel construction, the method comprising:

[0007] Multiple values ​​are taken within the preset range of spray parameters to obtain multiple spray parameter data; each value taken generates one set of spray parameter data.

[0008] For each spray parameter data, the spray parameter data and the current tunnel environment data are used as inputs, and the trained prediction model is used to obtain the predicted value of the tunnel dust concentration after treatment corresponding to the spray parameter data.

[0009] The spray parameter data corresponding to the minimum predicted value of the tunnel dust concentration after treatment is selected as the target parameter data, and dust treatment is carried out based on the target parameter data.

[0010] In some embodiments, the spray parameters include spray air pressure, spray water pressure, spray angle, and surfactant concentration; the tunnel environment data includes actual values ​​of the tunnel environment, which includes tunnel temperature, tunnel humidity, tunnel wind speed, and tunnel dust concentration before treatment.

[0011] In some embodiments, before obtaining the predicted value of the post-treatment tunnel dust concentration corresponding to the spray parameter data using a trained prediction model with the spray parameter data and the current tunnel environment data as input, the control method further includes: training a trained prediction model, specifically including:

[0012] Obtain a dataset; the dataset includes multiple samples and a label corresponding to each sample, the samples include sample spray parameter data and sample tunnel environment data, and the labels include the values ​​of tunnel dust concentration after sample treatment;

[0013] Construct an initial prediction model;

[0014] The initial prediction model is trained using the dataset to obtain a trained prediction model.

[0015] In some embodiments, the initial prediction model adopts a BP neural network model, including an input layer, a hidden layer and an output layer connected in sequence; the number of nodes in the hidden layer is determined according to an empirical formula.

[0016] In some embodiments, before training the initial prediction model using the dataset to obtain a trained prediction model, the control method further includes:

[0017] The dataset is normalized to obtain a normalized dataset, and the normalized dataset is used as a new dataset.

[0018] In some embodiments, before training the initial prediction model using the dataset to obtain a trained prediction model, the control method further includes:

[0019] The initial network parameters of the initial prediction model are determined by the hunter-prey optimization algorithm, and the parameter-initialized model is obtained. The parameter-initialized model is then used as the new initial prediction model.

[0020] In some embodiments, after dust control is performed based on the target parameter data, the control method further includes: obtaining the actual value of the tunnel dust concentration after control; using the target parameter data, the tunnel environment data, and the actual value of the tunnel dust concentration after control as a new sample, adding the new sample to the dataset to obtain an updated dataset; and retraining the trained prediction model using the updated dataset at preset intervals to obtain a retrained model, and using the retrained model as the trained prediction model for the next dust control operation.

[0021] A dust control system for tunnel construction, the control system comprising:

[0022] The value acquisition module is used to acquire multiple values ​​within a preset range of spray parameters to obtain multiple spray parameter data; each value acquisition generates one set of spray parameter data.

[0023] The prediction module is used to obtain the predicted value of the tunnel dust concentration after treatment corresponding to each spray parameter data by using the spray parameter data and the current tunnel environment data as input and a trained prediction model.

[0024] The treatment module is used to select the spray parameter data corresponding to the minimum predicted value of the tunnel dust concentration after treatment as the target parameter data, and to perform dust treatment based on the target parameter data.

[0025] A dust control device for tunnel construction includes:

[0026] Processor; and

[0027] Memory, in which computer-readable program instructions are stored.

[0028] The control method described above is executed when the computer-readable program instructions are run by the processor.

[0029] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described control method.

[0030] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0031] This invention provides a method, system, equipment, and medium for controlling dust during tunnel construction. It involves taking multiple values ​​of spray parameters within a preset range to obtain multiple spray parameter data. For each spray parameter data, using the spray parameter data and current tunnel environmental data as input, a trained prediction model is used to obtain a predicted value of the dust concentration in the tunnel after treatment. The spray parameter data corresponding to the minimum predicted value of the dust concentration after treatment is selected as the target parameter data. Dust control is performed based on the target parameter data. This allows for the selection of spray parameter data that minimizes the dust concentration in the tunnel after treatment, based on the tunnel environmental data. The spray parameters can be automatically and dynamically adjusted according to the actual construction site, improving dust suppression effectiveness, reducing dust suppression costs, and solving the problems of poor dust suppression effect, water waste, and high dust suppression costs caused by the inability to automatically adjust spray parameters according to the actual construction site. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a flowchart of the control method provided in Embodiment 1 of the present invention;

[0034] Figure 2 This is a system block diagram of the control system provided in Embodiment 2 of the present invention. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] The purpose of this invention is to provide a method, system, equipment, and medium for controlling dust during tunnel construction, which can automatically and dynamically adjust spray parameters according to the actual construction site, thereby improving dust suppression effect and reducing dust suppression costs.

[0037] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0038] Example 1:

[0039] This embodiment provides a method for controlling dust during tunnel construction, applicable to dust control during tunnel excavation and daily production operations. Figure 1 As shown, the control method includes:

[0040] S1: Multiple values ​​are taken within the preset range of spray parameters to obtain multiple spray parameter data; each value taken generates one set of spray parameter data.

[0041] S2: For each spray parameter data, using the spray parameter data and the current tunnel environment data as input, the predicted value of the tunnel dust concentration after treatment corresponding to the spray parameter data is obtained by using the trained prediction model;

[0042] S3: Select the spray parameter data corresponding to the minimum predicted value of the tunnel dust concentration after treatment as the target parameter data, and carry out dust treatment based on the target parameter data.

[0043] Dust refers to fine solid particles that can be suspended in the air, generated by natural or mechanical forces. Suspended particles with a diameter of less than 75 μm are internationally defined as dust. Dust from tunnel construction is mainly composed of silica dust. Spray dust suppression technology is commonly used in tunnel construction for dust control. The purpose of this embodiment is to determine the optimal spray parameters for dust suppression equipment using spray dust suppression technology. Spray dust suppression refers to the method and technology of dispersing water into droplets and spraying them towards the dust source to suppress and capture dust. Its principle is to use a high-pressure pump to pressurize water, which is then sent through a high-pressure pipeline to a high-pressure nozzle for atomization, forming a drifting water mist. Because the water mist particles are micron-sized and extremely fine, they can adsorb impurities in the air, creating a clean and fresh air environment, achieving multiple functions such as dust suppression and humidification. Depending on the dust suppression equipment used, the types of spray parameters will vary. In practical applications, users can determine the type of spray parameters based on the type of dust suppression equipment they are using. As an example, the dust suppression equipment used in this embodiment operates as follows: a surfactant is added to water, and a high-pressure pump is used to pressurize the water containing the surfactant and deliver it to a high-pressure nozzle. The high-pressure nozzle is simultaneously connected to high-pressure gas, which atomizes the water containing the surfactant within the nozzle, forming a water mist. This water mist is then sprayed at a certain angle by the high-pressure nozzle to the dust source, achieving the purpose of dust suppression. The surfactant ensures uniform distribution of the water mist, improving dust suppression efficiency. Based on the above operating process, the spray parameters in this embodiment include spray air pressure (i.e., the pressure of the high-pressure gas), spray water pressure (i.e., the pressure of the water containing the surfactant), spray angle (i.e., the angle at which the high-pressure nozzle sprays the water mist), and surfactant concentration.

[0044] After determining the category of spray parameters, this embodiment can perform multiple value selections within the preset value range of the spray parameters. Each value selection generates a spray parameter value, which is recorded as one spray parameter data point. Multiple value selections yield multiple spray parameter data points; that is, the spray parameter data in this embodiment includes one spray parameter value. Generally, dust suppression equipment will provide a value range for each category of spray parameters. This value range is the preset value range of the spray parameters described in this embodiment. If not provided, the preset value range can be determined based on experience. Specifically, when the spray parameters include spray air pressure, spray water pressure, spray angle, and surfactant concentration, selecting one value from each of the preset value ranges for spray air pressure, spray water pressure, spray angle, and surfactant concentration will generate one spray parameter value, resulting in one spray parameter data point. Repeating this process multiple times yields multiple spray parameter data points.

[0045] Preferably, when taking multiple values ​​within a preset range of spray parameters, the values ​​can be taken starting from the minimum value of the preset range and at preset intervals. Specifically, when the spray parameters include spray air pressure, spray water pressure, spray angle, and surfactant concentration, the values ​​can be taken multiple times starting from the minimum value of the preset range of spray air pressure at a first preset interval, starting from the minimum value of the preset range of spray water pressure at a second preset interval, starting from the minimum value of the preset range of spray angle at a third preset interval, and starting from the minimum value of the preset range of surfactant concentration at a fourth preset interval. The preset intervals for each category of spray parameters can be the same or different. The nth value of each category of spray parameters forms a value for the spray parameter, resulting in a spray parameter data.

[0046] The tunnel environment data in this embodiment includes actual values ​​of the tunnel environment. The category of the tunnel environment can be determined according to the actual situation of the tunnel, as long as it can characterize the tunnel environment. The actual value refers to the value of the tunnel environment collected by sensors. As an example, the tunnel environment in this embodiment includes tunnel temperature, tunnel humidity, tunnel wind speed, and tunnel dust concentration before treatment. When dust control is required, this embodiment can read the values ​​of tunnel temperature, tunnel humidity, tunnel wind speed, and tunnel dust concentration before treatment collected by sensors to obtain the current actual value of the tunnel environment. This current actual value of the tunnel environment is the current tunnel environment data.

[0047] Before using spray parameter data and current tunnel environment data as input to obtain the predicted value of tunnel dust concentration after treatment corresponding to the spray parameter data using a trained prediction model, the control method of this embodiment further includes the step of training a trained prediction model, which may include:

[0048] (1) Obtain the dataset, which includes multiple samples and labels for each sample. The samples include sample spray parameter data and sample tunnel environment data. The labels include the values ​​of tunnel dust concentration after sample treatment.

[0049] This embodiment can construct a database of factors influencing dust control during tunnel construction, and further select multiple sets of data from this database to form a dataset. The steps for constructing the database of factors influencing dust control during tunnel construction are as follows:

[0050] 1) Establish a model of influencing factors for dust control during tunnel construction, i.e., determine the data categories in the database. These categories mainly include spray parameters, tunnel environment, and post-treatment tunnel dust concentration. Spray parameters mainly include spray air pressure, spray water pressure, spray angle, and surfactant concentration. Tunnel environment mainly includes tunnel temperature, tunnel humidity, tunnel wind speed, and pre-treatment tunnel dust concentration. Both pre-treatment and post-treatment tunnel dust concentrations can be characterized by PM10 and PM2.5, thus each concentration has two possible values.

[0051] 2) A database of influencing factors for dust control during tunnel construction was constructed using a dust testing platform in the laboratory. Simulation experiments of spray atomization dust suppression under various changing conditions were conducted on the laboratory dust testing platform. These experiments involved altering tunnel temperature, humidity, wind speed, pre-treatment tunnel dust concentration, spray pressure, water pressure, spray angle, and surfactant concentration to simulate dust control. The dust concentration after dust suppression was then measured using a dust concentration meter. The resulting data (tunnel temperature, humidity, wind speed, pre-treatment tunnel dust concentration, spray pressure, water pressure, spray angle, surfactant concentration, and post-treatment tunnel dust concentration) constituted a set of data in the database. Repeating the simulation experiments multiple times yielded multiple sets of data.

[0052] (2) Construct the initial prediction model.

[0053] The initial prediction model in this embodiment can be any machine learning model. As an example, the initial prediction model in this embodiment uses a backpropagation (BP) neural network model. Considering that a three-layer neural network can approximate any complex nonlinear system, the initial prediction model in this embodiment includes an input layer, a hidden layer, and an output layer connected in sequence. The number of nodes in the input layer is the same as the number of data categories included in the samples in the dataset, specifically 8 in this embodiment. The number of nodes in the output layer is the same as the number of data categories included in the labels in the dataset, specifically 1 in this embodiment. The number of nodes in the hidden layer is determined according to an empirical formula, which is: Where l is the number of nodes in the hidden layer, m is the number of nodes in the input layer, n is the number of nodes in the output layer, and a is an adjustment constant with a value between [1, 10], which is an integer. Therefore, in this embodiment, the number of nodes in the input layer is 8, the number of nodes in the hidden layer is determined according to an empirical formula, and the number of nodes in the output layer is 1.

[0054] Preferably, this embodiment determines the optimal number of nodes in the hidden layer using a trial-and-error method. Specifically, the optimal number of nodes is determined by comparing the errors of models with different node counts. First, an empirical formula is used to determine the range of possible node counts. Then, for each value within this range, a specific initial prediction model is determined. This initial prediction model is trained and validated using a dataset to obtain the error corresponding to that value. The error can be expressed as the average relative error, which is the average difference between the predicted and actual values ​​for each sample used in validation. Finally, the optimal number of nodes in the hidden layer is determined by comparing the errors corresponding to different values. The value corresponding to the minimum error is the optimal number of nodes in the hidden layer. After testing, this embodiment found that the error is minimized when the number of nodes in the hidden layer is 10. Therefore, the optimal number of nodes in the hidden layer is 10, and the structure of the initial prediction model is determined to be 8-10-1.

[0055] (3) Use the dataset to train the initial prediction model to obtain the trained prediction model.

[0056] In this embodiment, tunnel temperature, tunnel humidity, tunnel wind speed, tunnel dust concentration before treatment, spray air pressure, spray water pressure, spray angle, and surfactant concentration are used as feature parameters, and tunnel dust concentration after treatment is used as label parameter. These parameters are input into the initial prediction model. After multiple iterations, once the model reaches convergence accuracy, the corresponding trained prediction model can be obtained.

[0057] The dimensions and units of data in different categories within a dataset often differ, frequently resulting in data values ​​not being on the same order of magnitude. This can lead to problems such as difficulty in convergence or gradient explosion during neural network training. Therefore, it is preferable to normalize the original data in the dataset. Thus, before training the initial prediction model using the dataset to obtain a trained prediction model, the control method in this embodiment further includes: normalizing the dataset to obtain a normalized dataset, and using the normalized dataset as the new dataset to perform the subsequent step of training the initial prediction model using the dataset.

[0058] This embodiment uses Min-Max normalization to map the original data to the range [0, 1]. Min-Max normalization, also known as deviation normalization, is a linear transformation of the original data, mapping the transformed values ​​to the range [0, 1]. The specific transformation function used for each category of data is as follows:

[0059]

[0060] Where, x new x represents the transformed data; x represents the original data; x represents the transformed data. min x is the minimum value of the original data. max This represents the maximum value of the original data.

[0061] Traditional neural networks are prone to getting stuck in local optima during training, leading to suboptimal network performance after training. To address this issue, before training the initial prediction model using a dataset to obtain a trained prediction model, the control method in this embodiment further includes: using the Hunter-Prey Optimization (HPO) algorithm to determine the initial network parameters of the initial prediction model, obtaining a parameter-initialized model, and using this parameter-initialized model as the new initial prediction model to execute the subsequent step of training the initial prediction model using the dataset. When the initial prediction model is a BP neural network model, the initial network parameters include initial weights and initial thresholds, specifically including the weights and biases from the input layer to the hidden layer and the weights and biases from the hidden layer to the output layer. This initial prediction model can be denoted as the HPO-BP neural network model.

[0062] The process of determining the initial network parameters of the initial prediction model using the hunter-prey optimization algorithm is as follows: (1) Generate an initial population, which includes multiple search agents.

[0063] The hunter-prey optimization algorithm randomly generates a certain number of search agents (divided into hunters and prey) within the search space. Each search agent is a vector group with a certain spatial dimension (representing the position of the search agent). In the optimization problem of the initial network parameters of the BP neural network, this vector group represents a potential optimal solution for the weights and thresholds. The spatial dimension of the vector group is determined by the total number of weights and thresholds to be optimized in the neural network. Based on the determined structure of the BP neural network, the spatial dimension N of the vector group can be calculated using N = n + l + m × l + n × l.

[0064] (2) Calculate the fitness of each search agent in the initial group to obtain the optimal fitness value of the initial group.

[0065] Constructing the fitness function is a crucial step in optimizing a backpropagation (BP) neural network using the hunter-prey optimization algorithm. The fitness function used in this embodiment is as follows:

[0066]

[0067] Where F is the fitness; N is the spatial dimension of the vector group; Y i y represents the predicted output value of the neural network. i The label value is the expected output value of the neural network. After determining the initial network parameters of the initial prediction model by searching for the agent's location, the initial prediction model is trained and validated using the dataset. The trained prediction model can then be used to determine the predicted output value of the neural network, which is then used to further calculate the fitness.

[0068] The fitness of each search agent is calculated using the above formula, and the minimum fitness value is selected as the optimal fitness value of the initial population.

[0069] (3) Update the initial population to obtain the updated population.

[0070] Search agents can be divided into hunters and prey. The update formula for each search agent in the initial group is as follows:

[0071]

[0072] Where, x i (t) represents the current position of the i-th hunter; C is the balance parameter; Z is the adaptive parameter; P pos(j) Let be the position of the j-th prey; μ(j) be the average position of all search agents in the initial swarm; R5 be a random number in the range [0, 1]; β be an adjustment parameter; T pos R1 represents the global optimal position, which is the position of the search agent corresponding to the optimal fitness value of the initial population; R2 is a random number in the range [-1, 1].

[0073] For each search agent in the initial population, a random number R5 is generated. It is determined whether R5 is less than the adjustment parameter β. If it is, the search agent is regarded as a hunter, and the position of the search agent is updated using the first hunter update formula of the update formula. If not, the search agent is regarded as prey, and the position of the search agent is updated using the second prey update formula of the update formula. This process is repeated to update the position of the entire initial population and obtain the updated population.

[0074] (4) Calculate the fitness of each search agent in the updated population to obtain the optimal fitness value of the updated population.

[0075] (5) Determine whether the iteration termination condition has been met; if yes, use the position of the search agent corresponding to the optimal fitness value of the updated population as the initial network parameter of the initial prediction model; if no, use the updated population as the initial population of the next iteration, use the optimal fitness value of the updated population as the optimal fitness value of the initial population of the next iteration, and return to the step of "updating the initial population to obtain the updated population".

[0076] The iteration termination condition in this embodiment can be reaching the maximum number of iterations.

[0077] This embodiment uses the hunter-prey optimization algorithm to optimize the initial threshold and initial weights of the BP neural network, and establishes the HPO-BP neural network model. This solves the problem that traditional neural networks are prone to getting trapped in local optima, and obtains a more efficient and accurate prediction model. That is, the HPO algorithm is used to optimize the initial weights and initial thresholds of the BP, which effectively improves the prediction accuracy of the neural network model.

[0078] Preferably, in order to set a reasonable number of search agents and iterations, this embodiment uses different numbers of search agents for comparative optimization experiments, and obtains the interval of iterations when the fitness converges for each number of agents and the fitness size after convergence, as shown in Table 1. The iteration count at convergence is presented as an interval because even after convergence, the fitness may fluctuate slightly, making it impossible to pinpoint a specific iteration number; an interval representation is more reasonable and accurate.

[0079] Table 1

[0080]

[0081] As shown in Table 1, the fitness value after convergence is minimized when the number of search agents is 200, indicating that this number of search agents provides the best optimization effect for the initial weights and thresholds of the BP neural network. Furthermore, the number of iterations required for fitness convergence falls within the range of [300, 500] for different numbers of search agents. Based on these results, to reduce the number of iterations while ensuring sufficient optimization of the neural network and thus improving computational efficiency, the number of search agents is set to 200 and the number of iterations to 500.

[0082] When dust control is required, the current tunnel environment data collected by on-site sensors and the data of each spray parameter are input into the trained prediction model. The trained prediction model is used to obtain the predicted value of the tunnel dust concentration after control for each spray parameter data. The predicted value of the tunnel dust concentration after control refers to the predicted tunnel dust concentration after dust control according to the spray parameter data in the current tunnel environment. The optimal value of the spray parameters is determined based on the predicted value of the tunnel dust concentration after control. Specifically, the spray parameter data corresponding to the minimum value of the predicted value of the tunnel dust concentration after control is selected as the target parameter data. The target parameter data includes the optimal spray air pressure, spray water pressure, spray angle and surfactant concentration. Dust control is carried out based on the target parameter data, that is, the dust suppression equipment is controlled to work according to the target parameter data.

[0083] In this embodiment, the tunnel dust control compliance value can also be preset. After obtaining the predicted value of the tunnel dust concentration after treatment corresponding to each spray parameter data, the predicted value of the tunnel dust concentration after treatment corresponding to each spray parameter data is compared with the tunnel dust control compliance value. If the predicted value of the tunnel dust concentration after treatment corresponding to each spray parameter data is greater than the tunnel dust control compliance value, then S1 is used again to obtain the value, so as to ensure that the dust control standard can be met after dust control based on the target parameter data.

[0084] This embodiment integrates the trained prediction model into the construction site dust control system. The construction site dust control system is connected to sensors and dust suppression equipment. The sensors transmit field data in real time, and predictions are made based on the field data. The resulting optimal dust concentration control parameters (i.e., target parameter data) are transmitted to the dust suppression equipment, which can complete the setting of the spray parameters of the dust suppression equipment. It can quickly and dynamically adjust the spray parameters to achieve automated and precise dust control.

[0085] This embodiment discloses a machine learning-based method for controlling dust in tunnel construction. It studies the factors influencing dust during tunnel construction, constructs a database of these factors, generates a dataset based on the database, and inputs the dataset into an initial prediction model based on a hunter-prey optimization algorithm. Tunnel environmental data (tunnel temperature, tunnel humidity, tunnel wind speed, and pre-treatment tunnel dust concentration) and spray parameter data (spray air pressure, spray water pressure, spray angle, and surfactant concentration) are used as feature parameters, and the post-treatment tunnel dust concentration is used as a label parameter. The initial prediction model is then trained to obtain a well-trained prediction model. This model can be further used to determine the optimal values ​​of spray parameters under certain tunnel conditions for better dust control. The control method of this embodiment is applicable to dust control in various tunnel excavation and daily production operations, reducing the workload of dust control personnel, achieving automated and precise dust control, improving dust control effectiveness, and saving dust control costs.

[0086] Preferably, due to the limited data in the database, this embodiment utilizes data from tunnel dust control to perform feedback control, forming a closed loop. Simultaneously with dust control, the completed data is updated to the database of factors influencing tunnel construction dust control. The neural network is iteratively trained at fixed intervals to evolve and improve prediction accuracy. Specifically, the actual dust concentration at the site is collected after each dust control operation based on spray parameters. The dust concentration after dust suppression at the construction site is then added back to the dataset for updating. A self-correcting function for the neural network model is introduced, using the updated dataset to correct the model, continuously improving its accuracy during use and achieving a closed loop for data prediction. This control method can be called a closed-loop control method. This closed-loop control method can correct the accuracy of the prediction model and can be widely applied to all tunnel excavation construction.

[0087] Therefore, after dust control based on target parameter data, the control method in this embodiment further includes: obtaining the actual value of the tunnel dust concentration after control; using the target parameter data, tunnel environment data, and the actual value of the tunnel dust concentration after control as a new sample; adding the new sample to the dataset to obtain an updated dataset. Every preset time interval, the trained prediction model is retrained using the updated dataset to obtain a retrained model, which is then used as the trained prediction model for the next dust control operation. It should be noted that after every preset time interval, the number of new samples in the updated dataset is the same as the number of dust control operations performed during that period.

[0088] This embodiment utilizes machine learning algorithms to fully consider the influencing factors of tunnel construction dust. It establishes the correlation between the effectiveness of fine dust control in tunnel construction and local environmental parameters, dust concentration, and spray parameters. Combined with theoretical data from laboratory simulations, the machine learning model is trained, resulting in a well-trained model that enables adaptive parameter adjustment for various types of fine dust from tunnel construction. This provides a foundation for intelligent adjustment of dynamic dust suppression system parameters in response to changing tunnel construction environments. Simultaneously, by incorporating positive feedback theory, the adjusted spray parameters, environmental parameters, and dust concentration before and after control are used as feedback values ​​and re-inputted into the database. Through periodic iterative updates to the neural network, this method significantly improves its adaptability to different tunnel types while maintaining high precision and intelligent dust suppression. This embodiment considers dynamic environmental changes in the fine dust control environment of tunnel construction. It uses machine learning methods combined with real-time dynamic data from on-site sensors to optimize and adjust spray parameters. The information feedback mechanism updates the machine learning model, improving environmental adaptability and achieving precise control of fine dust from tunnel construction while saving water resources and reducing dust suppression costs. The control method in this embodiment significantly reduces the dust concentration in the tunnel, effectively solving the problems of difficulty in dynamically adjusting parameters and poor concentration removal effect during dust suppression in tunnel construction. It is of great significance for maintaining the surrounding natural ecology and has certain reference value for dust pollution prevention and control in green construction processes.

[0089] Example 2:

[0090] This embodiment provides a dust control system for tunnel construction, such as... Figure 2 As shown, the control system includes:

[0091] The value acquisition module M1 is used to acquire multiple values ​​within a preset range of spray parameters to obtain multiple spray parameter data; each acquisition generates one set of spray parameter data.

[0092] The prediction module M2 is used to obtain the predicted value of the tunnel dust concentration after treatment corresponding to each spray parameter data by using the spray parameter data and the current tunnel environment data as input and a trained prediction model.

[0093] The treatment module M3 is used to select the spray parameter data corresponding to the minimum predicted value of the tunnel dust concentration after treatment as the target parameter data, and to perform dust treatment based on the target parameter data.

[0094] Example 3:

[0095] This embodiment provides a dust control device for tunnel construction, including:

[0096] Processor; and

[0097] Memory, in which computer-readable program instructions are stored.

[0098] The control method described in Embodiment 1 is executed when the computer-readable program instructions are run by the processor.

[0099] Example 4:

[0100] This embodiment provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the control method described in Embodiment 1.

[0101] Each embodiment in this specification focuses on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be found in the method section.

[0102] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for controlling dust during tunnel construction, characterized in that, The control method includes: Multiple values ​​are taken within the preset range of spray parameters to obtain multiple spray parameter data; each value taken generates one set of spray parameter data. For each spray parameter data, the spray parameter data and the current tunnel environment data are used as inputs, and the trained prediction model is used to obtain the predicted value of the tunnel dust concentration after treatment corresponding to the spray parameter data. If the predicted value of the tunnel dust concentration after treatment corresponding to each of the spray parameter data is greater than the tunnel dust treatment standard value, then the value is taken again; otherwise, the spray parameter data corresponding to the minimum value of the predicted value of the tunnel dust concentration after treatment is selected as the target parameter data, and dust treatment is carried out based on the target parameter data. Before using the spray parameter data and current tunnel environment data as input to obtain the predicted value of the tunnel dust concentration after treatment corresponding to the spray parameter data using a trained prediction model, the control method further includes: training the trained prediction model, specifically including: Obtain a dataset; the dataset includes multiple samples and a label corresponding to each sample, the samples include sample spray parameter data and sample tunnel environment data, and the labels include the values ​​of tunnel dust concentration after sample treatment; An initial prediction model is constructed. The initial prediction model adopts a BP neural network model, including an input layer, a hidden layer, and an output layer connected in sequence. The number of nodes in the hidden layer is determined according to an empirical formula. Specifically, the range of values ​​for the number of nodes in the hidden layer is determined by the empirical formula. For each value within the range, a specific initial prediction model is determined. The specific initial prediction model is trained and validated using a dataset to obtain the error corresponding to the value. The value corresponding to the minimum error is selected as the optimal value for the number of nodes in the hidden layer. The initial network parameters of the initial prediction model are determined using a hunter-prey optimization algorithm. Specifically, an initial population is generated, comprising multiple search agents. The fitness of each search agent in the initial population is calculated, based on the predicted and expected output values ​​of the BP neural network model. The optimal fitness value of the initial population is obtained. The initial population is then updated to obtain an updated population. The fitness of each search agent in the updated population is calculated to obtain the optimal fitness value of the updated population. This process continues until the iteration termination condition is met. The position of the search agent corresponding to the optimal fitness value of the updated population is used as the initial network parameters of the initial prediction model, resulting in a parameter-initialized model. This parameter-initialized model is then used as the new initial prediction model. The iteration termination condition is reaching the maximum number of iterations. The initial prediction model is trained using the dataset to obtain a trained prediction model; Comparative optimization experiments were conducted using different numbers of search agents to obtain the range of iterations required for convergence of each number of fitness agents and the fitness value after convergence, thereby determining the number of search agents and the number of iterations.

2. The control method according to claim 1, characterized in that, The spray parameters include spray air pressure, spray water pressure, spray angle, and surfactant concentration; the tunnel environment data includes actual values ​​of the tunnel environment, including tunnel temperature, tunnel humidity, tunnel wind speed, and tunnel dust concentration before treatment.

3. The control method according to claim 1, characterized in that, Before training the initial prediction model using the dataset to obtain a trained prediction model, the control method further includes: The dataset is normalized to obtain a normalized dataset, and the normalized dataset is used as a new dataset.

4. The control method according to claim 1, characterized in that, After dust control is performed based on the target parameter data, the control method further includes: obtaining the actual value of the tunnel dust concentration after control; using the target parameter data, the tunnel environment data, and the actual value of the tunnel dust concentration after control as a new sample, adding the new sample to the dataset to obtain an updated dataset; and retraining the trained prediction model using the updated dataset at preset intervals to obtain a retrained model, and using the retrained model as the trained prediction model for the next dust control operation.

5. A dust control system for tunnel construction, characterized in that, The control system includes: The value acquisition module is used to acquire multiple values ​​within a preset range of spray parameters to obtain multiple spray parameter data; each value acquisition generates one set of spray parameter data. The prediction module is used to obtain the predicted value of the tunnel dust concentration after treatment corresponding to each spray parameter data by using the spray parameter data and the current tunnel environment data as input and a trained prediction model. The treatment module is used to take a value again if the predicted value of the tunnel dust concentration after treatment corresponding to each of the spray parameter data is greater than the tunnel dust treatment compliance value; otherwise, the spray parameter data corresponding to the minimum value of the predicted value of the tunnel dust concentration after treatment is selected as the target parameter data, and dust treatment is carried out based on the target parameter data. Before using the spray parameter data and current tunnel environment data as input to obtain the predicted value of the tunnel dust concentration after treatment corresponding to the spray parameter data using a trained prediction model, the control system further includes: training the trained prediction model, specifically including: Obtain a dataset; the dataset includes multiple samples and a label corresponding to each sample, the samples include sample spray parameter data and sample tunnel environment data, and the labels include the values ​​of tunnel dust concentration after sample treatment; An initial prediction model is constructed. The initial prediction model adopts a BP neural network model, including an input layer, a hidden layer, and an output layer connected in sequence. The number of nodes in the hidden layer is determined according to an empirical formula. Specifically, the range of values ​​for the number of nodes in the hidden layer is determined by the empirical formula. For each value within the range, a specific initial prediction model is determined. The specific initial prediction model is trained and validated using a dataset to obtain the error corresponding to the value. The value corresponding to the minimum error is selected as the optimal value for the number of nodes in the hidden layer. The initial network parameters of the initial prediction model are determined using a hunter-prey optimization algorithm. Specifically, an initial population is generated, comprising multiple search agents. The fitness of each search agent in the initial population is calculated, based on the predicted and expected output values ​​of the BP neural network model. The optimal fitness value of the initial population is obtained. The initial population is then updated to obtain an updated population. The fitness of each search agent in the updated population is calculated to obtain the optimal fitness value of the updated population. This process continues until the iteration termination condition is met. The position of the search agent corresponding to the optimal fitness value of the updated population is used as the initial network parameters of the initial prediction model, resulting in a parameter-initialized model. This parameter-initialized model is then used as the new initial prediction model. The iteration termination condition is reaching the maximum number of iterations. The initial prediction model is trained using the dataset to obtain a trained prediction model; Comparative optimization experiments were conducted using different numbers of search agents to obtain the range of iterations required for convergence of each number of fitness agents and the fitness value after convergence, thereby determining the number of search agents and the number of iterations.

6. A dust control device for tunnel construction, characterized in that, include: processor; as well as Memory, in which computer-readable program instructions are stored. The control method described in any one of claims 1-4 is executed when the computer-readable program instructions are run by the processor.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the control method according to any one of claims 1-4.

Images