1. A shield constructs hobbing cutter wearing and tearing monitoring devices based on ballast piece appearance, its characterized in that includes:

the data acquisition box is arranged on one side of the shield machine conveying belt;

the industrial personal computer is arranged on one side of the data acquisition box, the first end of the industrial personal computer is electrically connected with the first end of the data acquisition box, and the second end of the industrial personal computer is electrically connected with the computer;

the first end of the truss is erected on the shield tunneling machine conveyor belt, and the second end of the truss is erected on the ground;

the two illuminating devices are symmetrically arranged at two ends of the truss and are detachably connected with the truss;

the camera is arranged at the top end of the truss, is detachably connected with the truss, and is electrically connected with the second end of the data acquisition box;

the infrared thermometer is arranged at the top end of the truss, the infrared thermometer is detachably connected with the truss, and the infrared thermometer is electrically connected with the third end of the data acquisition box.

2. The shield hob abrasion monitoring method based on the ballast flake morphology is applied to the shield hob abrasion monitoring device based on the ballast flake morphology as claimed in claim 1, and is characterized by comprising the following steps of:

step 1, obtaining a known shield engineering case in a computer and constructing a database;

step 2, analyzing the mapping relation between hob structure parameters, engineering geological parameters, shield tunneling parameters, slag sheet morphology parameters and temperature parameters in a database and the abrasion loss of the hob by adopting a GRNN neural network to obtain a mapping relation analysis result;

step 3, constructing a hob abrasion prediction system based on the database, the GRNN neural network and the mapping relation analysis result and setting a hob abrasion alarm threshold;

step 4, collecting hob structure parameters and engineering geological parameters before the construction of the current shield machine and inputting the hob abrasion parameters into a hob abrasion prediction system;

step 5, collecting shield tunneling parameters, slag sheet shape parameters and temperature parameters in real time during construction of the current shield tunneling machine and inputting the parameters into a hob abrasion prediction system;

step 6, the hob abrasion prediction system predicts the hob abrasion loss according to the input hob structure parameters, engineering geological parameters, shield tunneling parameters, slag sheet shape parameters and temperature parameters and compares the predicted hob abrasion loss with a set hob abrasion loss alarm threshold;

step 7, when the predicted hob abrasion loss exceeds the hob abrasion loss alarm threshold, the hob abrasion loss prediction system gives an alarm, the shield machine stops running, the hob is replaced, the abrasion loss of the abraded hob is measured and input into the hob abrasion loss prediction system;

step 8, performing self-adaptive optimization on the GRNN neural network by adopting a particle swarm optimization algorithm according to the wear loss of the worn hob and the predicted hob wear loss;

step 9, restarting the shield machine after tool changing is completed, skipping to step 4 to continue execution until the whole shield engineering work is completed, and finishing monitoring;

and 10, when the predicted hob abrasion loss does not exceed the hob abrasion loss alarm threshold, skipping to the step 5 to continue execution until the whole shield engineering work is finished, and finishing monitoring.

3. The shield hob abrasion monitoring method based on the ballast fragment morphology as claimed in claim 2, wherein the step 1 specifically includes:

the shield engineering case comprises a plurality of cutter changing records, wherein the cutter changing records comprise hob structure parameters, engineering geological parameters, shield tunneling parameters, ballast sheet morphology parameters, temperature parameters and hob abrasion loss, the hob structure parameters comprise cutting edge structure parameters, hob radius, cutter spacing and load weight, the engineering geological parameters comprise rock mechanics parameters, rock material parameters, rock joint and fault parameters and lithology parameters, the shield tunneling parameters comprise penetration degree of a shield machine, rotating speed of a cutter head of the shield machine, thrust of the shield machine, working torque of the shield machine, soil bin pressure of the shield machine and propelling speed of the shield machine, the ballast sheet morphology parameters comprise ballast sheet particle size distribution indexes, ballast sheet length and minor axis ratio indexes and ballast sheet texture indexes, and the temperature parameters comprise ballast sheet temperature.

4. The shield hob abrasion monitoring method based on the ballast fragment morphology as claimed in claim 3, wherein the step 2 specifically includes:

taking the penetration degree, the propulsion speed, the cutter head rotating speed, the shield machine thrust, the shield machine working torque, the shield machine soil bin pressure and the hob radius of a shield machine as 7 neurons of an input layer of the GRNN neural network, and taking the cutter abrasion loss as the neurons of an output layer to form the GRNN neural network;

dividing data in a database into 100 groups, selecting 10 groups of data from 100 groups of data as a test set by adopting a random sampling method, and taking the remaining 90 groups of data as a training set;

randomly dividing a training set into 9 units, wherein each unit comprises 10 groups of data, randomly selecting 8 units from the 9 units as training set input samples by adopting a cross validation method, using the remaining 1 unit as the training set output samples, normalizing the training set input sample data to be between [ -1,1], carrying out validation search by using step length 0.01 in (0,1], searching for a smoothing factor sigma which enables the mean square error of a predicted value and a sample value to be minimum, and recording an optimal input sample and an optimal output sample corresponding to the current smoothing factor;

and normalizing the data of the test set, and constructing a 4-layer GRNN neural network by taking the obtained smoothing factor sigma, the optimal input sample and the optimal output sample as input variables, and outputting the tool wear loss by an output layer.

5. The shield hob abrasion monitoring method based on the ballast fragment morphology as claimed in claim 4, wherein the step 3 specifically includes:

based on a mapping relation between engineering geological parameters, shield tunneling parameters, a shield hob structure, slag sheet feature sizes and temperature parameters and cutter abrasion loss in a database are analyzed and collected through correlation by a GRNN neural network, a hob abrasion prediction system is established:

δ_i＝β₀p+β₁v+β₂n+β₃F+β₄T+β₅S+β₆L+······+β_mX_m+C (1)

wherein, beta₀、β₁、β₂、β₃、β₄····β_mIs the parameter to be estimated, δ_iFor the abrasion amount of the cutter, p represents the penetration degree of the shield machine, v represents the propulsion speed, n represents the rotation speed of a cutter head, F represents the thrust force of the shield machine, T represents the working torque of the shield machine, L represents the radius of a hob, S represents the size of the slag fragments, and X represents the abrasion amount of the cutter_mOther parameters related to the wear amount of the tool are shown, and C represents a undetermined constant.

6. The shield hob abrasion monitoring method based on the ballast fragment morphology as claimed in claim 5, wherein the step 5 specifically includes:

the method comprises the steps of acquiring shield tunneling parameters in real time when the current shield tunneling machine is constructed through a main control room of the shield tunneling machine and inputting the shield tunneling parameters into a hob abrasion prediction system, shooting pictures of ballast pieces on a shield tunneling machine conveying belt in real time through a camera and inputting the pictures into a computer to calculate the size of the ballast pieces, obtaining the ballast piece appearance parameters and inputting the ballast piece appearance parameters into the hob abrasion prediction system, and measuring the temperature of the ballast pieces on the shield tunneling machine conveying belt in real time through an infrared thermometer and inputting the temperature into the hob abrasion prediction system.

7. The shield hob abrasion monitoring method based on the ballast fragment morphology as claimed in claim 6, wherein the step 8 specifically includes:

optimizing the CRNN neural network by adopting a particle swarm optimization algorithm: setting particle swarm calculation parameters, setting the mean square error of the cutter abrasion loss predicted by the hob abrasion prediction system when an alarm is given and the abrasion loss of the abraded hob as a fitness function, substituting a learning sample and an example into a GRNN neural network, and calculating a fitness value F_iComparing the fitness values of all positions passed by the ith particle to determine the optimal position P_biComparing all particles at their optimal positions P_biDetermining the optimal position G of the whole population_bAdjusting the speed and position of the particles according to the self position and the optimal position of each particleWhen the iteration termination condition is reached, the optimal position G is obtained_bUsing the searched optimal position G_bAnd optimizing the GRNN neural network.

Background

With the wide application of tunnel boring machines at home and abroad, shield construction has gradually become a main construction method in tunnel construction. For shield construction, shield hobbing cutters are used as the main rock breaking tool. The material is suitable for various rock stratums, gravel stratums and soft and hard composite stratums and is widely used in various heading machines.

Due to the uncertainty of rock and the complexity of stratum in the tunneling process, the dynamic property of the stress change of the cutter and the uncertainty of the abrasion form of the cutter are caused. After the long-time cutting of rock, the hob needs to be replaced when abnormal abrasion or normal abrasion occurs to a certain degree, so that the shield tunneling efficiency and the construction cost are inevitably influenced.

Practice shows that the shape and size of rock fragments and the particle size distribution thereof are indirect feedback of complex surrounding rock conditions and mechanical tunneling performance, and the wear state of a cutter can be predicted through fragment parameters and shield operation parameters. Currently, the current practice is. Whether the tool needs to be put into a bin for changing is judged only by the experience of constructors, the judgment accuracy is different from person to person, and serious accidents are likely to happen due to judgment errors. Therefore, a shield hob abrasion monitoring method based on the ballast fragment morphology needs to be provided.

Disclosure of Invention

The invention provides a shield hob abrasion monitoring device and method based on ballast piece morphology, and aims to solve the problems that the abrasion loss of a current hob cannot be accurately predicted by a traditional construction method, and the intelligent, safe and efficient tunneling requirements of a shield cannot be met.

In order to achieve the above object, an embodiment of the present invention provides a shield hob abrasion monitoring device based on ballast fragment morphology, including:

the data acquisition box is arranged on one side of the shield machine conveying belt;

the industrial personal computer is arranged on one side of the data acquisition box, the first end of the industrial personal computer is electrically connected with the first end of the data acquisition box, and the second end of the industrial personal computer is electrically connected with the computer;

the first end of the truss is erected on the shield tunneling machine conveyor belt, and the second end of the truss is erected on the ground;

the two illuminating devices are symmetrically arranged at two ends of the truss and are detachably connected with the truss;

the camera is arranged at the top end of the truss, is detachably connected with the truss, and is electrically connected with the second end of the data acquisition box;

the infrared thermometer is arranged at the top end of the truss, the infrared thermometer is detachably connected with the truss, and the infrared thermometer is electrically connected with the third end of the data acquisition box.

The embodiment of the invention also provides a shield hob abrasion monitoring method based on the ballast fragment morphology, which comprises the following steps:

step 1, obtaining a known shield engineering case in a computer and constructing a database;

step 2, analyzing the mapping relation between hob structure parameters, engineering geological parameters, shield tunneling parameters, slag sheet morphology parameters and temperature parameters in a database and the abrasion loss of the hob by adopting a GRNN neural network to obtain a mapping relation analysis result;

step 3, constructing a hob abrasion prediction system based on the database, the GRNN neural network and the mapping relation analysis result and setting a hob abrasion alarm threshold;

step 4, collecting hob structure parameters and engineering geological parameters before the construction of the current shield machine and inputting the hob abrasion parameters into a hob abrasion prediction system;

step 5, collecting shield tunneling parameters, slag sheet shape parameters and temperature parameters in real time during construction of the current shield tunneling machine and inputting the parameters into a hob abrasion prediction system;

step 6, the hob abrasion prediction system predicts the hob abrasion loss according to the input hob structure parameters, engineering geological parameters, shield tunneling parameters, slag sheet shape parameters and temperature parameters and compares the predicted hob abrasion loss with a set hob abrasion loss alarm threshold;

step 7, when the predicted hob abrasion loss exceeds the hob abrasion loss alarm threshold, the hob abrasion loss prediction system gives an alarm, the shield machine stops running, the hob is replaced, the abrasion loss of the abraded hob is measured and input into the hob abrasion loss prediction system;

step 8, performing self-adaptive optimization on the GRNN neural network by adopting a particle swarm optimization algorithm according to the wear loss of the worn hob and the predicted hob wear loss;

step 9, restarting the shield machine after tool changing is completed, skipping to step 4 to continue execution until the whole shield engineering work is completed, and finishing monitoring;

and 10, when the predicted hob abrasion loss does not exceed the hob abrasion loss alarm threshold, skipping to the step 5 to continue execution until the whole shield engineering work is finished, and finishing monitoring.

Wherein, the step 1 specifically comprises:

the shield engineering case comprises a plurality of cutter changing records, wherein the cutter changing records comprise hob structure parameters, engineering geological parameters, shield tunneling parameters, ballast sheet morphology parameters, temperature parameters and hob abrasion loss, the hob structure parameters comprise cutting edge structure parameters, hob radius, cutter spacing and load weight, the engineering geological parameters comprise rock mechanics parameters, rock material parameters, rock joint and fault parameters and lithology parameters, the shield tunneling parameters comprise penetration degree of a shield machine, rotating speed of a cutter head of the shield machine, thrust of the shield machine, working torque of the shield machine, soil bin pressure of the shield machine and propelling speed of the shield machine, the ballast sheet morphology parameters comprise ballast sheet particle size distribution indexes, ballast sheet length and minor axis ratio indexes and ballast sheet texture indexes, and the temperature parameters comprise ballast sheet temperature.

Wherein, the step 2 specifically comprises:

taking the penetration degree, the propulsion speed, the cutter head rotating speed, the shield machine thrust, the shield machine working torque, the shield machine soil bin pressure and the hob radius of a shield machine as 7 neurons of an input layer of the GRNN neural network, and taking the cutter abrasion loss as the neurons of an output layer to form the GRNN neural network;

dividing data in a database into 100 groups, selecting 10 groups of data from 100 groups of data as a test set by adopting a random sampling method, and taking the remaining 90 groups of data as a training set;

randomly dividing a training set into 9 units, wherein each unit comprises 10 groups of data, randomly selecting 8 units from the 9 units as training set input samples by adopting a cross validation method, using the remaining 1 unit as the training set output samples, normalizing the training set input sample data to be between [ -1,1], carrying out validation search by using step length 0.01 in (0,1], searching for a smoothing factor sigma which enables the mean square error of a predicted value and a sample value to be minimum, and recording an optimal input sample and an optimal output sample corresponding to the current smoothing factor;

and normalizing the data of the test set, and constructing a 4-layer GRNN neural network by taking the obtained smoothing factor sigma, the optimal input sample and the optimal output sample as input variables, and outputting the tool wear loss by an output layer.

Wherein, the step 3 specifically comprises:

based on a mapping relation between engineering geological parameters, shield tunneling parameters, a shield hob structure, slag sheet feature sizes and temperature parameters and cutter abrasion loss in a database are analyzed and collected through correlation by a GRNN neural network, a hob abrasion prediction system is established:

δ_i＝β₀p+β₁v+β₂n+β₃F+β₄T+β₅S+β₆L+······+β_mX_m+C (1)

wherein, beta₀、β₁、β₂、β₃、β₄····β_mIs the parameter to be estimated, δ_iFor the abrasion amount of the cutter, p represents the penetration degree of the shield machine, v represents the propulsion speed, n represents the rotation speed of a cutter head, F represents the thrust force of the shield machine, T represents the working torque of the shield machine, L represents the radius of a hob, S represents the size of the slag fragments, and X represents the abrasion amount of the cutter_mOther parameters related to the wear amount of the tool are shown, and C represents a undetermined constant.

Wherein, the step 5 specifically comprises:

the method comprises the steps of acquiring shield tunneling parameters in real time when the current shield tunneling machine is constructed through a main control room of the shield tunneling machine and inputting the shield tunneling parameters into a hob abrasion prediction system, shooting pictures of ballast pieces on a shield tunneling machine conveying belt in real time through a camera and inputting the pictures into a computer to calculate the size of the ballast pieces, obtaining the ballast piece appearance parameters and inputting the ballast piece appearance parameters into the hob abrasion prediction system, and measuring the temperature of the ballast pieces on the shield tunneling machine conveying belt in real time through an infrared thermometer and inputting the temperature into the hob abrasion prediction system.

Wherein, the step 8 specifically comprises:

optimizing the CRNN neural network by adopting a particle swarm optimization algorithm: setting particle swarm calculation parameters, setting the mean square error of the cutter abrasion loss predicted by the hob abrasion prediction system when an alarm is given and the abrasion loss of the abraded hob as a fitness function, substituting a learning sample and an example into a GRNN neural network, and calculating a fitness value F_iComparing the fitness values of all positions passed by the ith particle to determine the optimal position P_biComparing all particles at their optimal positions P_biDetermining the optimal position G of the whole population_bAdjusting the speed and position of the particles according to the self position of each particle and the optimal particle position, and obtaining the optimal position G when the iteration termination condition is reached_bUsing the searched optimal position G_bAnd optimizing the GRNN neural network.

The scheme of the invention has the following beneficial effects:

according to the shield hob abrasion monitoring device and method based on the ballast piece morphology, the abrasion state of the hob can be predicted in real time according to actual engineering geological parameters, hob structural parameters, real-time tunneling parameters and ballast piece parameters, an alarm function is provided when the predicted abrasion state of the hob reaches a set threshold value, and accurate prediction of hob abrasion can be improved through closed-loop feedback adaptive adjustment according to different projects.

Drawings

FIG. 1 is a flow chart of the present invention;

FIG. 2 is a schematic structural view of the present invention;

FIG. 3 is a schematic diagram of a CRNN neural network of the present invention;

fig. 4 is a flowchart of the particle swarm optimization algorithm of the present invention.

[ description of reference ]

1-shield machine conveyor belt; 2-a data collection box; 3-an industrial personal computer; 4-a truss; 5-a lighting device; 6-a camera; 7-infrared thermometer.

Detailed Description

In order to make the technical problems, technical solutions and advantages of the present invention more apparent, the following detailed description is given with reference to the accompanying drawings and specific embodiments.

The invention provides a shield hob abrasion monitoring device and method based on ballast fragment morphology, aiming at the problems that the abrasion loss of the current hob cannot be accurately predicted by the existing construction method, and the intelligent, safe and efficient shield tunneling requirements cannot be met.

As shown in fig. 1 to 4, an embodiment of the present invention provides a shield hob abrasion monitoring device based on ballast fragment morphology, including: the data acquisition box 2 is arranged on one side of the shield machine conveyor belt 1; the industrial personal computer 3 is arranged on one side of the data acquisition box 2, a first end of the industrial personal computer 3 is electrically connected with a first end of the data acquisition box 2, and a second end of the industrial personal computer is electrically connected with a computer; the first end of the truss 4 is erected on the shield tunneling machine conveyor belt 1, and the second end of the truss 4 is erected on the ground; two lighting devices 5 are arranged, the two lighting devices 5 are symmetrically arranged at two ends of the truss 4, and the two lighting devices 5 are detachably connected with the truss 4; the camera 6 is arranged at the top end of the truss 4, the camera 6 is detachably connected with the truss 4, and the camera 6 is electrically connected with the second end of the data acquisition box 2; the infrared thermometer 7 is arranged at the top end of the truss 4, the infrared thermometer 7 is detachably connected with the truss 4, and the infrared thermometer 7 is electrically connected with the third end of the data acquisition box 2.

According to the shield hob abrasion monitoring device and method based on the ballast morphology, the truss 4 is used for supporting the illuminating device 5, the camera 6 and the infrared thermometer 7, the illuminating device 5 is used for being matched with the camera 6 to perform image datamation, and the infrared camera 6 is used for measuring the temperature of the ballast in real time.

The embodiment of the invention also provides a shield hob abrasion monitoring method based on the ballast fragment morphology, which comprises the following steps: step 1, obtaining a known shield engineering case in a computer and constructing a database; step 2, analyzing the mapping relation between hob structure parameters, engineering geological parameters, shield tunneling parameters, slag sheet morphology parameters and temperature parameters in a database and the abrasion loss of the hob by adopting a GRNN neural network to obtain a mapping relation analysis result; step 3, constructing a hob abrasion prediction system based on the database, the GRNN neural network and the mapping relation analysis result and setting a hob abrasion alarm threshold; step 4, collecting hob structure parameters and engineering geological parameters before the construction of the current shield machine and inputting the hob abrasion parameters into a hob abrasion prediction system; step 5, collecting shield tunneling parameters, slag sheet shape parameters and temperature parameters in real time during construction of the current shield tunneling machine and inputting the parameters into a hob abrasion prediction system; step 6, the hob abrasion prediction system predicts the hob abrasion loss according to the input hob structure parameters, engineering geological parameters, shield tunneling parameters, slag sheet shape parameters and temperature parameters and compares the predicted hob abrasion loss with a set hob abrasion loss alarm threshold; step 7, when the predicted hob abrasion loss exceeds the hob abrasion loss alarm threshold, the hob abrasion loss prediction system gives an alarm, the shield machine stops running, the hob is replaced, the abrasion loss of the abraded hob is measured and input into the hob abrasion loss prediction system; step 8, performing self-adaptive optimization on the GRNN neural network by adopting a particle swarm optimization algorithm according to the wear loss of the worn hob and the predicted hob wear loss; step 9, restarting the shield machine after tool changing is completed, skipping to step 4 to continue execution until the whole shield engineering work is completed, and finishing monitoring; and 10, when the predicted hob abrasion loss does not exceed the hob abrasion loss alarm threshold, skipping to the step 5 to continue execution until the whole shield engineering work is finished, and finishing monitoring.

According to the shield hob abrasion monitoring device and method based on the slag piece morphology, a mapping relation between collected parameters and hob abrasion loss is analyzed through a CRNN neural network, a hob abrasion loss alarm threshold value is set according to engineering tunneling data and the fluctuation change condition of the slag piece morphology, abnormal hob abrasion state characteristics are predicted in advance, and various hob failure conditions such as normal abrasion, eccentric abrasion and the like are accurately identified; the main process of prediction and judgment by adopting the neural network is as follows: 1. selecting a neural network; 2. selecting a learning sample; 3. determining an input vector and an output vector; 4. setting parameters of a neural network; 5. learning a neural network, and constructing a neural network model; 6. inputting a target object for prediction judgment; using a GRNN neural network structure, wherein the GRNN neural network structure comprises an input layer, a mode layer, a summation layer and an output layer, and training the GRNN neural network structure by using a part of existing data; the number of neurons in the input layer is the number of input vectors in the learning sample, the input layer outputs a plurality of input variables to the mode layer, the number of neurons in the mode layer is equal to the number of input vectors in the learning sample, each neuron in the mode layer corresponds to different input vectors, and the neuron transfer function of the mode layer is as follows:

wherein, P_iRepresents the output of the ith neuron of the pattern layer, X represents the learning sample, and X ═ X₁,X₂,…,X_n]^T，X_aRepresents the a input vector in the learning sample, sigma represents the smoothing factor, and i represents the i neuron.

Two types are used in the summation layer to sum the outputs of the neurons of the mode layer:

one type of calculation is:

arithmetically summing the output of each neuron of the mode layer, wherein the connection weight of the mode layer and each neuron is 1, and the neuron transfer function of the summation layer is as follows:

the second category of calculation methods is:

carrying out weighted summation on each neuron of the mode layer, wherein the connection weight value between the ith neuron in the mode layer and the jth numerator summation neuron in the summation layer is the ith output sample Y_iThe ith element in (1), the summation layer neuron transfer function, is as follows:

the number of neurons in the output layer is equal to the dimension k of the output vector in the learning sample, each neuron divides the output of the summation layer, the output of neuron j, as follows:

according to the shield hob abrasion monitoring device and method based on the ballast piece morphology, parameters such as engineering geology, a hob structure and a hob head structure are input into a system as known parameters before construction, excavation parameters and ballast piece morphology parameters of a shield machine are monitored in real time in the construction process and input into the system, real-time abrasion loss of a hob is predicted, when the abrasion loss of the hob reaches a set hob abrasion loss alarm threshold value, an alarm is given to prompt the hob to be changed, a worker enters a bin to change the hob after shutdown and measures the current abrasion state and abrasion loss of the hob, the result is input into a hob abrasion prediction system, feedback adaptive adjustment of the hob abrasion prediction system is carried out, the prediction robustness of the hob abrasion prediction system is improved under the condition of high prediction accuracy of the current engineering, and the cycle is repeated after the hob is changed each time.

Wherein, the step 1 specifically comprises: the shield engineering case comprises a plurality of cutter changing records, wherein the cutter changing records comprise hob structure parameters, engineering geological parameters, shield tunneling parameters, ballast sheet morphology parameters, temperature parameters and hob abrasion loss, the hob structure parameters comprise cutting edge structure parameters, hob radius, cutter spacing and load weight, the engineering geological parameters comprise rock mechanics parameters, rock material parameters, rock joint and fault parameters and lithology parameters, the shield tunneling parameters comprise penetration degree of a shield machine, rotating speed of a cutter head of the shield machine, thrust of the shield machine, working torque of the shield machine, soil bin pressure of the shield machine and propelling speed of the shield machine, the ballast sheet morphology parameters comprise ballast sheet particle size distribution indexes, ballast sheet length and minor axis ratio indexes and ballast sheet texture indexes, and the temperature parameters comprise ballast sheet temperature.

Wherein, the step 2 specifically comprises: taking the penetration degree, the propulsion speed, the cutter head rotating speed, the shield machine thrust, the shield machine working torque, the shield machine soil bin pressure and the hob radius of a shield machine as 7 neurons of an input layer of the GRNN neural network, and taking the cutter abrasion loss as the neurons of an output layer to form the GRNN neural network;

dividing data in a database into 100 groups, selecting 10 groups of data from 100 groups of data as a test set by adopting a random sampling method, and taking the remaining 90 groups of data as a training set;

randomly dividing a training set into 9 units, wherein each unit comprises 10 groups of data, randomly selecting 8 units from the 9 units as training set input samples by adopting a cross validation method, using the remaining 1 unit as the training set output samples, normalizing the training set input sample data to be between [ -1,1], carrying out validation search by using step length 0.01 in (0,1], searching for a smoothing factor sigma which enables the mean square error of a predicted value and a sample value to be minimum, and recording an optimal input sample and an optimal output sample corresponding to the current smoothing factor;

and normalizing the data of the test set, and constructing a 4-layer GRNN neural network by taking the obtained smoothing factor sigma, the optimal input sample and the optimal output sample as input variables, and outputting the tool wear loss by an output layer.

Wherein, the step 3 specifically comprises: based on a mapping relation between engineering geological parameters, shield tunneling parameters, a shield hob structure, slag sheet feature sizes and temperature parameters and cutter abrasion loss in a database are analyzed and collected through correlation by a GRNN neural network, a hob abrasion prediction system is established:

δ_i＝β₀p+β₁v+β₂n+β₃F+β₄T+β₅S+β₆L+······+β_mX_m+C (1)

wherein, beta₀、β₁、β₂、β₃、β₄····β_mIs the parameter to be estimated, δ_iFor the abrasion amount of the cutter, p represents the penetration degree of the shield machine, v represents the propulsion speed, n represents the rotation speed of a cutter head, F represents the thrust force of the shield machine, T represents the working torque of the shield machine, L represents the radius of a hob, S represents the size of the slag fragments, and X represents the abrasion amount of the cutter_mOther parameters related to the wear amount of the tool are shown, and C represents a undetermined constant.

Wherein, the step 5 specifically comprises: the method comprises the steps of acquiring shield tunneling parameters in real time when the current shield tunneling machine is constructed through a main control room of the shield tunneling machine and inputting the shield tunneling parameters into a hob abrasion prediction system, shooting pictures of ballast pieces on a shield tunneling machine conveyor belt 1 in real time through a camera 6 and inputting the pictures into a computer to calculate the size of the ballast pieces, obtaining the ballast piece morphology parameters and inputting the ballast piece morphology parameters into the hob abrasion prediction system, and measuring the temperature of the ballast pieces on the shield tunneling machine conveyor belt 1 in real time through an infrared thermometer 7 and inputting the temperature into the hob abrasion prediction system.

In the shield hob abrasion monitoring device and method based on the ballast piece morphology, the camera 6, the infrared thermometer 7, the data acquisition box 2 and the industrial personal computer 3 are connected to transmit and store image data and temperature data to the hob abrasion prediction system in the computer through data lines, and the current hob abrasion condition is analyzed in real time through the hob abrasion prediction system.

Wherein, the step 8 specifically comprises: optimizing the CRNN neural network by adopting a particle swarm optimization algorithm: setting particle swarm calculation parameters, setting the mean square error of the cutter abrasion loss predicted by the hob abrasion prediction system when an alarm is given and the abrasion loss of the abraded hob as a fitness function, substituting a learning sample and an example into a GRNN neural network, and calculating a fitness value F_iComparing the fitness values of all positions passed by the ith particle to determine the optimal position P_biComparing all particles at their optimal positions P_biDetermining the optimal position G of the whole population_bAdjusting the speed and position of the particles according to the self position of each particle and the optimal particle position, and obtaining the optimal position G when the iteration termination condition is reached_bUsing the searched optimal position G_bAnd optimizing the GRNN neural network.

The shield hob abrasion monitoring device and the shield hob abrasion monitoring method based on the ballast piece morphology, which are disclosed by the embodiment of the invention, are characterized in that before the shield machine is constructed, the shield hob abrasion monitoring device is built on the shield machine conveyor belt 1, the hob structure parameters and the engineering geological parameters of the current shield machine are manually collected and input into the hob abrasion prediction system, when the shield machine runs, the method comprises the steps of acquiring slag sheet shape parameters and temperature parameters in the running process of a shield machine in real time through a shield hob abrasion monitoring device, inputting the parameters and the temperature parameters into a hob abrasion prediction system, specifically, shooting pictures of slag sheets on a conveyor belt 1 of the shield machine in real time through a camera 6, inputting the pictures into a computer to calculate the size of the slag sheet shape, inputting the acquired slag sheet shape parameters into the hob abrasion prediction system, and storing the parameters into the hob abrasion prediction system, measuring the temperature of the slag pieces on the shield machine conveyor belt 1 in real time through the infrared thermometer 7, inputting the temperature and storing the temperature into the hob abrasion prediction system; acquiring shield tunneling parameters input in the current shield tunneling machine construction process in real time through a main control room of the shield tunneling machine and storing the shield tunneling parameters input into the hob abrasion prediction system; the hob abrasion prediction system outputs the hob abrasion loss corresponding to the current input parameters according to the input hob structure parameters, engineering geological parameters, shield tunneling parameters, slag sheet shape parameters, temperature parameters and the existing mapping relation analysis result, judges whether the output hob abrasion loss exceeds a set hob abrasion loss alarm threshold value or not, and continues the hob abrasion loss prediction of the next group of parameters when the output hob abrasion loss does not exceed the set hob abrasion loss alarm threshold value; when the output hob abrasion loss exceeds a set hob abrasion loss alarm threshold value, an alarm is given, the shield machine is stopped, a worker replaces a hob of the shield machine, the hob abrasion loss of the current abraded hob is measured and input into the hob abrasion loss prediction system for storage, the hob abrasion loss of the current abraded hob and the covariance of the predicted hob abrasion loss are used as a fitness function of a particle swarm optimization algorithm to find an optimal smooth factor, a CRNN neural network is optimized according to the optimal smooth factor, the shield machine is started, the hob abrasion loss prediction of the next group of parameters is continued until the whole shield engineering work is completed, and monitoring is finished.

The shield hob abrasion monitoring device and the shield hob abrasion monitoring method based on the slag piece morphology, which are disclosed by the embodiment of the invention, are characterized in that actual engineering geological parameters, hob structural parameters, real-time tunneling parameters and slag piece parameters are acquired and input into a hob abrasion prediction system, the hob abrasion prediction system can predict the abrasion state of a hob in real time, the hob abrasion prediction system provides an alarm function when the predicted abrasion state of the hob reaches a set hob abrasion amount alarm threshold value, when the hob abrasion prediction system alarms, the GRNN neural network parameters are optimized by adopting a particle swarm optimization algorithm according to the predicted hob abrasion amount and the actual hob abrasion amount, the influence of human factors on the neural network design is reduced, the GRNN neural network is adaptively improved, an adaptive improvement algorithm is added on the basis of the original mapping relationship, and under the condition of keeping high prediction accuracy, the robustness of hob abrasion prediction is improved, so that the shield hob abrasion monitoring device and method based on the ballast piece morphology can improve the accurate prediction of hob abrasion through closed-loop feedback self-adaptive adjustment according to different projects.

While the foregoing is directed to the preferred embodiment of the present invention, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention as defined in the appended claims.