CN116519055A - An intelligent early warning method for pipeline blockage of mud-water shield mud-water circulation system - Google Patents

Abstract

The invention discloses an intelligent pre-warning method for pipeline blockage of a mud-water shield mud-water circulation system. The steps are as follows: firstly, using a detection device to detect the state of the flow field in the pipeline and the movement trajectory of particles, and using coupling simulation to calculate model performance parameters , together with the detection results of the detection device to construct a training data set. Secondly, construct the proxy model and substitute the training sample set into the proxy model for training. During the training process, the optimization algorithm is used to optimize the weight threshold of the proxy model, and the accuracy of the model is evaluated. After meeting the accuracy requirements, the model is saved and predicted. The control system according to The prediction results are adjusted and given an early warning, so as to achieve the effect of intelligent early warning of the mud-water circulation system pipeline and improve the excavation efficiency.

Description

An intelligent early warning method for pipeline blockage in mud-water shield mud-water circulation system

technical field

The invention belongs to the technical field of mud-water shield intelligent monitoring, and in particular relates to an intelligent pre-warning method for pipeline blockage in a mud-water circulation system of a mud-water shield.

Background technique

With the acceleration of urbanization and the continuous expansion of transportation infrastructure construction, mud-water shields have also been widely used. Among them, the mud-water circulation system is an important part of the mud-water balance shield machine, and the mud particles are transported out by pipeline transportation. , so as to realize the recycling of mud and water. The configuration of the mud and water circulation system determines the construction efficiency of the shield machine, and its working stability is of great significance to the efficient tunneling and safe construction of the shield.

Since the mud carries solid particles of different components such as gravel and pebbles, the mud-water circulation system is prone to the accumulation of particles in the pipeline and the problem of pipeline blockage. Once the pipeline is clogged, it will have a serious impact on the mud-water circulation system, which may lead to excessive pressure in the pipeline, pipe burst and other consequences, resulting in abnormal shutdown of the shield tunneling work, which will eventually reduce construction efficiency and increase costs. Among the currently existing pipeline blockage detection methods, the detection method based on detection equipment to detect pipelines has high efficiency, but the use cost is relatively high. Therefore, how to realize the intelligent early warning method for pipeline blockage in the mud-water shield mud-water circulation system can be monitored in real time It is of great significance to reduce energy consumption and improve tunneling efficiency by timely taking corresponding adjustment measures according to the operation status of the slurry circulation system.

Contents of the invention

Aiming at the problem that the construction of mud-water circulation pipelines is complicated and the pipelines are prone to particle clogging, the present invention provides a mud-water shield mud-water circulation system pipeline clogging intelligence with scientific principle, easy operation, high degree of intelligence, and low operating cost. Early warning method.

In order to solve the above-mentioned technical problems, the present invention adopts the following technical scheme: an intelligent early warning method for pipeline blockage of mud-water shield mud-water circulation system, comprising the following steps,

S1 uses the detection device to detect the state of the flow field in the pipeline and the trajectory of the particles, and displays the detection results on the visual interface;

S2 Carry out multi-physics simulation calculation according to the operating parameter information of the mud-water shield mud-water circulation system under real working conditions collected by the detection device, and integrate the multi-physics simulation calculation results to build a data set;

S3 will form a training sample set and substitute it into the proxy model for training. After the training is completed, test and use the verification sample set to verify the test results, and evaluate the prediction accuracy of the proxy model; if the accuracy meets the requirements, the motion state of the particle Predict and output the prediction results; if the accuracy does not meet the requirements, use the optimization algorithm to optimize the weights, and re-predict until the requirements are met, save the network and output the prediction results;

S4 transmits the predicted results to the control system, and the control system judges and adjusts the processing according to the predicted results, so as to achieve real-time monitoring of the pipelines of the mud-water circulation system of the mud-water shield.

Further, the detection device is X-ray beam, γ-ray or ultrasonic, and the detection device is used to detect the state of the flow field in the pipeline. Taking ultrasonic detection as an example, the detection device is installed at the entrance of the slurry feeding pipeline to detect the movement of particles State, speed, size, quantity and other parameters, as well as the electromagnetic flowmeter to measure the flow velocity in the pipeline, the specific detection steps are as follows:

S1.1 The ultrasonic transducer is installed on one side of the pipeline, the ultrasonic driving circuit excites the ultrasonic transducer to generate ultrasonic waves, and the ultrasonic waves enter the fluid;

S1.2 The echo signal receiving circuit completes the processing of the ultrasonic echo reflection signal, including signal isolation, amplification, filtering, and amplitude adjustment;

S1.3 Use the pulse refraction method to detect the state of the particles in the pipeline. Assuming that the propagation speed of the ultrasonic wave in the object is v, and the time from emission to the reflection of the particles is t, the sound path l of the particles is:

The received signal is displayed in the form of pulses on the display screen, and the position of the particles is judged according to the amplitude and time in the display screen. The number of particles can be directly determined by the number of pulse waves reflected by the particles, and t ₀ and t ₁ are used respectively Calculate the distribution position of the particles at all times, calculate the particle velocity, and display the detection results on the visual interface;

S1.4 Install an electromagnetic flowmeter, a pressure sensor, a density sensor and a viscosity sensor on the pipeline to measure the flow velocity, pressure, density and viscosity in the pipeline respectively;

S1.5 displays the detection results on the visual interface.

Further, the specific steps of step S2 are:

S2.1 Establish a calculation model according to the shape of the pipeline through Creo modeling software;

S2.2 Import the calculation model into the ICEM software grid division software, perform grid division on the established calculation model, check the quality of the grid, and export the .mesh grid file after the quality meets the requirements;

S2.3 Import the .mesh grid file into the EDEM software, and set the parameters of the particle motion model according to the collected particle motion state parameters;

S2.4 Import the .mesh grid file into the Fluent software, and set the boundary conditions of the simulation calculation model in combination with the actual working conditions; through iteration, calculate the momentum and energy exchange between the fluid and the particles, and convert the calculated The data is passed into EDEM;

S2.5 In the EDEM software, according to the flow field information calculated by the Fluent software, calculate the velocity and position information of the particles, and transfer the calculation results to Fluent for the next step of calculation;

S2.6 Repeat the above-mentioned coupling calculation process until the particle motion state tends to be stable. According to the simulation results, the performance parameters such as the slag-carrying efficiency and particle accumulation of the pipeline are obtained;

S2.7 Form the monitored sample points into a set of sample points, which can be expressed as {X ₁ , X ₂ , X ₃ , X ₄ , X ₅ , X ₆ } where X ₁ represents the velocity of the particle, and X ₂ represents the size of the particle , _X3 represents the size of the particle, _X4 represents the velocity of the fluid, _X5 and _X6 represent the viscosity and density of the mud, respectively.

Further, in step S3, the proxy model is used to predict the state of particle motion, and an optimization algorithm is used to optimize the predicted results. Taking the BP neural network as an example, these sample point sets are used as the input factors of the neural network. The specific implementation steps as follows:

S3.1 According to the Kolmogrov theorem, a 3-layer neural network with n input units 2n+1 intermediate units and m output units can accurately represent any mapping, and can coordinate the capacity of the intermediate layer with the training time; The structure of pipeline blockage prediction based on BP neural network is divided into 3 layers, 6 input layers, 13 hidden layers, and 3 output layers, which is called 6-13-3 network structure;

S3.2 There will be differences in the size of the sample points selected. In order to avoid the smaller data being submerged by the larger data, normalization processing is required to normalize the data to 0~1. The normalization formula is:

In the formula: X is the input factor, X _min is the minimum value of the data, and X _max is the maximum value of the data;

S3.3 According to the processed training data set, 200 data sample data sets are randomly selected, and the formed 200 sample data sets are used as the input nodes of the BP neural network, and the results of coupling simulation are output as the training set sample set to train the neural network ; Randomly select 40 groups of samples as the test set, test and use the verification set to verify the test results;

S3.4 Evaluate the model accuracy of neural network prediction:

S3.5 Use the GA genetic algorithm to optimize the weight and threshold. Each individual in the population includes a weight and threshold. The individual calculates the fitness through the fitness function. The GA genetic algorithm finds the most For the individual corresponding to the optimal applicability, the BP neural network re-predicts the weights and thresholds optimized by the genetic algorithm; after the training converges, the average flow velocity and total particle number of the slurry pipeline particles are predicted for the predicted samples.

Further, in step S4, the control system judges and adjusts according to the prediction result, specifically: if the prediction may cause blockage, send a signal warning and control the front-end device of the shield machine to adjust to avoid blockage; continue to detect after the adjustment is completed; If it is predicted that no blockage will occur, continue to detect at the next moment;

The specific measures for adjustment and treatment are as follows:

S4.1 Configure suitable high-quality mud, adjust the storage space of the ground mud, and adjust the viscosity of the mud; configure foam bentonite and inject it into the mud-water circulation system of the mud-water shield to improve the flow plasticity of the muck;

S4.2 Adjust the ratio and flow rate of the mud to maintain the proper performance of the mud; detect and treat the mud to remove impurities and sediments and maintain the cleanliness of the mud; check and maintain the equipment and pipelines in the mud recovery system to ensure It's working fine.

Adopt above-mentioned technical scheme, main beneficial effect of the present invention is as follows:

This method uses real data under various working conditions to perform multi-physics simulation calculations, uses simulation calculation results to train and predict pipeline proxy models, and uses optimization algorithms to optimize prediction results, thereby improving prediction accuracy. The control system judges the prediction results and whether to alarm, so as to realize the real-time monitoring of the blockage of the mud-water circulation system of the mud-water shield. Compared with the traditional detection method based on detection equipment to detect pipelines, this method can efficiently detect the movement of particles in the pipeline of the mud-water shield mud-water circulation system, perform real-time monitoring, reduce monitoring and maintenance costs, and improve construction efficiency.

Description of drawings

Fig. 1 is the overall flowchart of the present invention;

Figure 2 is a schematic diagram of measuring the state of the pipeline flow field using the pulse reflection method;

Fig. 3 is the coupling simulation flowchart of the present invention;

Fig. 4 is a flow chart of neural network prediction in the present invention.

Detailed ways

As shown in Figure 1, an intelligent pre-warning method for pipeline blockage of a mud-water shield mud-water circulation system includes the following steps:

S1 uses the detection device to detect the state of the flow field in the pipeline and the trajectory of the particles, and displays the detection results on the visual interface;

S2 Carry out multi-physics simulation calculation according to the operating parameter information of the mud-water shield mud-water circulation system under real working conditions collected by the detection device, and integrate the multi-physics simulation calculation results to build a data set;

S3 will form a training sample set and substitute it into the proxy model for training. After the training is completed, test and use the verification sample set to verify the test results, and evaluate the prediction accuracy of the proxy model; if the accuracy meets the requirements, the motion state of the particle Predict and output the prediction results; if the accuracy does not meet the requirements, use the optimization algorithm to optimize the weights, and re-predict until the requirements are met, save the network and output the prediction results;

S4 transmits the predicted results to the control system, and the control system judges and adjusts the processing according to the predicted results, so as to achieve real-time monitoring of the pipelines of the mud-water circulation system of the mud-water shield.

The detection device is X-ray beam, gamma ray or ultrasonic, and the detection device is used to detect the state of the flow field in the pipeline, as shown in Figure 2, taking ultrasonic detection as an example, the detection device is installed at the entrance of the slurry feeding pipeline, and the detection The movement state, speed, size, quantity and other parameters of the particles, as well as the electromagnetic flowmeter to measure the flow velocity in the pipeline, the specific detection steps are as follows:

S1.1 The ultrasonic transducer is installed on one side of the pipeline, the ultrasonic driving circuit excites the ultrasonic transducer to generate ultrasonic waves, and the ultrasonic waves enter the fluid;

S1.2 The echo signal receiving circuit completes the processing of the ultrasonic echo reflection signal, including signal isolation, amplification, filtering, and amplitude adjustment;

S1.3 Use the pulse refraction method to detect the state of the particles in the pipeline. Assuming that the propagation speed of the ultrasonic wave in the object is v, and the time from emission to the reflection of the particles is t, the sound path l of the particles is:

The received signal is displayed in the form of pulses on the display screen, and the position of the particles is judged according to the amplitude and time in the display screen. The number of particles can be directly determined by the number of pulse waves reflected by the particles, and t ₀ and t ₁ are used respectively Calculate the distribution position of the particles at all times, calculate the particle velocity, and display the detection results on the visual interface;

S1.4 Install an electromagnetic flowmeter, a pressure sensor, a density sensor and a viscosity sensor on the pipeline to measure the flow velocity, pressure, density and viscosity in the pipeline respectively;

S1.5 displays the detection results on the visual interface.

As shown in Figure 3, the specific steps of step S2 are:

S2.1 Establish a calculation model according to the shape of the pipeline through Creo modeling software;

S2.2 Import the calculation model into the ICEM software grid division software, perform grid division on the established calculation model, check the quality of the grid, and export the .mesh grid file after the quality meets the requirements;

S2.3 Import the .mesh grid file into the EDEM software, and set the parameters of the particle motion model according to the collected particle motion state parameters;

S2.4 Import the .mesh grid file into the Fluent software, and set the boundary conditions of the simulation calculation model in combination with the actual working conditions; through iteration, calculate the momentum and energy exchange between the fluid and the particles, and convert the calculated The data is passed into EDEM;

S2.5 In the EDEM software, according to the flow field information calculated by the Fluent software, calculate the velocity and position information of the particles, and transfer the calculation results to Fluent for the next step of calculation;

S2.6 Repeat the above-mentioned coupling calculation process until the particle motion state tends to be stable. According to the simulation results, the performance parameters such as the slag-carrying efficiency and particle accumulation of the pipeline are obtained;

S2.7 Form the monitored sample points into a set of sample points, which can be expressed as {X ₁ , X ₂ , X ₃ , X ₄ , X ₅ , X ₆ } where X ₁ represents the velocity of the particle, and X ₂ represents the size of the particle , _X3 represents the size of the particle, _X4 represents the velocity of the fluid, _X5 and _X6 represent the viscosity and density of the mud, respectively.

As shown in Figure 4, in step S3, the proxy model is used to predict the particle motion state, and the optimization algorithm is used to optimize the predicted results. Taking the BP neural network as an example, these sample point sets are used as the input factors of the neural network. The specific implementation steps are as follows:

S3.1 According to the Kolmogrov theorem, a 3-layer neural network with n input units 2n+1 intermediate units and m output units can accurately represent any mapping, and can coordinate the capacity of the intermediate layer with the training time; The structure of pipeline blockage prediction based on BP neural network is divided into 3 layers, 6 input layers, 13 hidden layers, and 3 output layers, which is called 6-13-3 network structure;

S3.2 There will be differences in the size of the sample points selected. In order to avoid the smaller data being submerged by the larger data, normalization processing is required to normalize the data to 0~1. The normalization formula is:

In the formula: X is the input factor, X _min is the minimum value of the data, and X _max is the maximum value of the data;

S3.3 According to the processed training data set, 200 data sample data sets are randomly selected, and the formed 200 sample data sets are used as the input nodes of the BP neural network, and the results of coupling simulation are output as the training set sample set to train the neural network ; Randomly select 40 groups of samples as the test set, test and use the verification set to verify the test results;

S3.4 Evaluate the model accuracy of neural network prediction:

S3.5 Use the GA genetic algorithm to optimize the weight and threshold. Each individual in the population includes a weight and threshold. The individual calculates the fitness through the fitness function. The GA genetic algorithm finds the most For the individual corresponding to the optimal applicability, the BP neural network re-predicts the weights and thresholds optimized by the genetic algorithm; after the training converges, the average flow velocity and total particle number of the slurry pipeline particles are predicted for the predicted samples.

In step S4, the control system judges and adjusts the processing according to the prediction results, specifically: if the prediction may cause blockage, send a signal warning and control the front-end device of the shield machine to adjust to avoid blockage; continue to detect after the adjustment is completed; if the prediction is not If there will be blockage, continue to detect at the next moment;

The specific measures for adjustment and treatment are as follows:

S4.1 Configure suitable high-quality mud, adjust the storage space of the ground mud, and adjust the viscosity of the mud; configure foam bentonite and inject it into the mud-water circulation system of the mud-water shield to improve the flow plasticity of the muck;

S4.2 Adjust the ratio and flow rate of the mud to maintain the proper performance of the mud; detect and treat the mud to remove impurities and sediments and maintain the cleanliness of the mud; check and maintain the equipment and pipelines in the mud recovery system to ensure It's working fine.

The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still understand the foregoing embodiments Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present invention shall be included in the scope of protection of the present invention. Inside.

Claims (5)

1. A mud-water shield mud-water circulation system pipeline blockage intelligent early warning method, is characterized in that: comprise the following steps, S1 uses the detection device to detect the state of the flow field in the pipeline and the trajectory of the particles, and displays the detection results on the visual interface; S2 Carry out multi-physics simulation calculation according to the operating parameter information of the mud-water shield mud-water circulation system under real working conditions collected by the detection device, and integrate the multi-physics simulation calculation results to build a data set; S3 will form a training sample set and substitute it into the proxy model for training. After the training is completed, test and use the verification sample set to verify the test results, and evaluate the prediction accuracy of the proxy model; if the accuracy meets the requirements, the motion state of the particle Predict and output the prediction results; if the accuracy does not meet the requirements, use the optimization algorithm to optimize the weights, and re-predict until the requirements are met, save the network and output the prediction results; S4 transmits the predicted results to the control system, and the control system judges and adjusts the processing according to the predicted results, so as to achieve real-time monitoring of the pipelines of the mud-water circulation system of the mud-water shield. 2. The intelligent early warning method for pipeline blockage of a muddy water shield muddy water circulation system according to claim 1, characterized in that: the detection device is X-ray beam, gamma ray or ultrasonic wave, and the detection device is used to detect the flow field in the pipeline Take ultrasonic detection as an example, the detection device is installed at the entrance of the slurry inlet pipeline to detect the movement state, speed, size, quantity and other parameters of the particles, and the electromagnetic flowmeter measures the flow velocity in the pipeline. The specific detection steps are as follows : S1.1 The ultrasonic transducer is installed on one side of the pipeline, the ultrasonic driving circuit excites the ultrasonic transducer to generate ultrasonic waves, and the ultrasonic waves enter the fluid; S1.2 The echo signal receiving circuit completes the processing of the ultrasonic echo reflection signal, including signal isolation, amplification, filtering, and amplitude adjustment; S1.3 Use the pulse refraction method to detect the state of the particles in the pipeline. Assuming that the propagation speed of the ultrasonic wave in the object is v, and the time from emission to the reflection of the particles is t, the sound path l of the particles is: The received signal is displayed in the form of pulses on the display screen, and the position of the particles is judged according to the amplitude and time in the display screen. The number of particles can be directly determined by the number of pulse waves reflected by the particles, and t ₀ and t ₁ are used respectively Calculate the distribution position of the particles at all times, calculate the particle velocity, and display the detection results on the visual interface; S1.4 Install an electromagnetic flowmeter, a pressure sensor, a density sensor and a viscosity sensor on the pipeline to measure the flow velocity, pressure, density and viscosity in the pipeline respectively; S1.5 displays the detection results on the visual interface. 3. The intelligent early warning method for pipeline blockage of a mud-water shield mud-water circulation system according to claim 2, characterized in that: the specific steps of step S2 are: S2.1 Establish a calculation model according to the shape of the pipeline through Creo modeling software; S2.2 Import the calculation model into the ICEM software grid division software, perform grid division on the established calculation model, check the quality of the grid, and export the .mesh grid file after the quality meets the requirements; S2.3 Import the .mesh grid file into the EDEM software, and set the parameters of the particle motion model according to the collected particle motion state parameters; S2.4 Import the .mesh grid file into the Fluent software, and set the boundary conditions of the simulation calculation model in combination with the actual working conditions; through iteration, calculate the momentum and energy exchange between the fluid and the particles, and convert the calculated The data is passed into EDEM; S2.5 In the EDEM software, according to the flow field information calculated by the Fluent software, calculate the velocity and position information of the particles, and transfer the calculation results to Fluent for the next step of calculation; S2.6 Repeat the above-mentioned coupling calculation process until the particle motion state tends to be stable. According to the simulation results, the performance parameters such as the slag-carrying efficiency and particle accumulation of the pipeline are obtained; S2.7 Form the monitored sample points into a set of sample points, which can be expressed as {X ₁ , X ₂ , X ₃ , X ₄ , X ₅ , X ₆ } where X ₁ represents the velocity of the particle, and X ₂ represents the size of the particle , _X3 represents the size of the particle, _X4 represents the velocity of the fluid, _X5 and _X6 represent the viscosity and density of the mud, respectively. 4. The intelligent pre-warning method for pipeline blockage of a mud-water shield mud-water circulation system according to claim 3, characterized in that: in step S3, a proxy model is used to predict the state of particle motion, and an optimization algorithm is used to predict the predicted results. For optimization, take the BP neural network as an example, and use these sample point sets as the input factors of the neural network. The specific implementation steps are as follows: S3.1 According to the Kolmogrov theorem, a 3-layer neural network with n input units 2n+1 intermediate units and m output units can accurately represent any mapping, and can coordinate the capacity of the intermediate layer with the training time; The structure of pipeline blockage prediction based on BP neural network is divided into 3 layers, 6 input layers, 13 hidden layers, and 3 output layers, which is called 6-13-3 network structure; S3.2 There will be differences in the size of the sample points selected. In order to avoid the smaller data being submerged by the larger data, normalization processing is required to normalize the data to 0~1. The normalization formula is: In the formula: X is the input factor, X _min is the minimum value of the data, and X _max is the maximum value of the data; S3.3 According to the processed training data set, 200 data sample data sets are randomly selected, and the formed 200 sample data sets are used as the input nodes of the BP neural network, and the results of coupling simulation are output as the training set sample set to train the neural network ; Randomly select 40 groups of samples as the test set, test and use the verification set to verify the test results; S3.4 Evaluate the model accuracy of neural network prediction: S3.5 Use the GA genetic algorithm to optimize the weight and threshold. Each individual in the population includes a weight and threshold. The individual calculates the fitness through the fitness function. The GA genetic algorithm finds the most For the individual corresponding to the optimal applicability, the BP neural network re-predicts the weights and thresholds optimized by the genetic algorithm; after the training converges, the average flow velocity and total particle number of the slurry pipeline particles are predicted for the predicted samples. 5. The intelligent pre-warning method for pipeline blockage of mud-water shield mud-water circulation system according to claim 4, characterized in that: in step S4, the control system judges and adjusts the processing according to the prediction result, specifically: if it is predicted that blockage may occur , send a signal warning and control the front-end device of the shield to adjust to avoid blockage; continue to detect after the adjustment is completed; if no blockage is predicted, continue to detect at the next moment; The specific measures for adjustment and treatment are as follows: S4.1 Configure suitable high-quality mud, adjust the storage space of the ground mud, and adjust the viscosity of the mud; configure foam bentonite and inject it into the mud-water circulation system of the mud-water shield to improve the flow plasticity of the muck; S4.2 Adjust the ratio and flow rate of the mud to maintain the proper performance of the mud; detect and treat the mud to remove impurities and sediments and maintain the cleanliness of the mud; check and maintain the equipment and pipelines in the mud recovery system to ensure It's working fine.