Intelligent cutting control method and system for heading machine

Abstract

The invention discloses an intelligent cutting control method and system for a heading machine, and the method comprises the following steps: collecting a feedback signal of the heading machine in a cutting process in real time through a sensor; calculating time domain, frequency domain and entropy value multi-domain characteristic parameters of the collected feedback signal to generate a characteristic vector; inputting the feature vector into a pre-constructed deep belief network model for hardness identification, and outputting the hardness of the current cut coal rock; according to the hardness, a preset cutting motion parameter database is retrieved, and a parameter combination corresponding to the hardness of the current cut coal rock is obtained; based on the obtained parameter combination, cooperatively controlling the rotating speed of a cutting head and the swing speed of a cutting arm of the heading machine; therefore, the coal rock hardness change is accurately sensed, the cutting parameters are automatically adjusted to be in the optimal state of the comprehensive performance, and the function of enhancing the self-adaptive capacity and the working stability under the complex coal seam condition is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent control of tunneling equipment, and in particular to an intelligent cutting control method and system for a tunneling machine. BACKGROUND

[0002] The cantilever type tunneling machine is the key equipment for tunneling in the coal mine, and its cutting performance directly affects the tunnel forming quality, tunneling efficiency and equipment service life. The current coal tunneling operation mainly relies on the experience of the operator for manual control, and faces the following technical bottlenecks:

[0003] Firstly, the underground coal seam has complex occurrence conditions, and the cutting conditions are variable, with a large range of coal and rock hardness (from less than 15MPa soft coal to more than 80MPa hard rock). It is difficult for the operator to accurately judge the change of coal and rock hardness in real time, which leads to improper selection of cutting parameters, and thus causes low cutting efficiency, high energy consumption, abnormal wear of cutting teeth and other problems, resulting in low work efficiency.

[0004] Secondly, although there are some automatic control schemes in the prior art, such as the patent with publication number CN113006793A which proposes an intelligent cutting joint control system for a cantilever type tunneling machine, which identifies the load through a neural network and controls the cutting head rotation speed and cutting arm swing speed by fuzzy reasoning. However, this scheme has obvious deficiencies: the control strategy uses simple fuzzy reasoning and step adjustment, with limited precision, and cannot guarantee that the tunneling machine always works in the optimal state in terms of comprehensive performance.

[0005] In addition, most of the existing technologies only adjust the cutting arm swing speed or only adjust the cutting head rotation speed, and lack a cooperative control mechanism between the two. When the cutting arm swing speed is adjusted in a large range, it is difficult to ensure that the tunneling machine is in the best working state with small cutting specific energy consumption, low dust production and small dynamic load of the transmission system.

[0006] Therefore, there is an urgent need for an intelligent system and method that can accurately identify the hardness of coal and rock and cooperatively and adaptively control the cutting head rotation speed and cutting arm swing speed based on optimized parameters. SUMMARY

[0007] The technical problem solved by the present application is the technical problem of low cutting efficiency, poor working condition adaptability and poor comprehensive cutting performance of the existing cantilever type tunneling machine under complex coal seam conditions.

[0008] To solve the above technical problems, the present application provides the following technical solutions:

[0009] An intelligent cutting control method for a tunneling machine, comprising the following steps:

[0010] Step S100, collecting feedback signals of the roadheader in the cutting process in real time through a sensor; the feedback signals include a current signal of a cutting motor, a vibration acceleration signal of a cutting arm and a driving torque signal of a hydraulic cylinder;

[0011] Step S200, calculating time domain, frequency domain and entropy value multi-domain feature parameters of the collected feedback signals to generate a feature vector;

[0012] Step S300, inputting the feature vector into a pre-constructed deep belief network model for hardness identification and outputting hardness of the current cutting coal rock;

[0013] Step S400, searching a preset cutting motion parameter database according to the hardness to obtain a parameter combination corresponding to the hardness of the current cutting coal rock; wherein the parameter combination includes a recommended rotating speed of a cutting head and a recommended swing speed of a cutting arm;

[0014] Step S500, performing collaborative control on the rotating speed of the cutting head and the swing speed of the cutting arm of the roadheader based on the obtained parameter combination.

[0015] Preferably, the calculation of the time domain, frequency domain and entropy value multi-domain feature parameters of the collected feedback signals includes:

[0016] calculating an amplitude mean value, an amplitude variance, a peak-to-peak value, a waveform factor, a kurtosis factor, a pulse factor, a permutation entropy, a singularity entropy, a sample entropy, a wavelet packet energy, a 4-layer wavelet packet singular value and a 6-layer wavelet packet singular value of the current signal of the cutting motor;

[0017] calculating an amplitude mean value, an amplitude variance, an amplitude kurtosis, a barycentric frequency, a mean square frequency, a variance frequency, a permutation entropy, a singularity entropy, a sample entropy, a wavelet packet energy, a 4-layer wavelet packet singular value and a 6-layer wavelet packet singular value of the vibration acceleration signal of the cutting arm;

[0018] calculating an amplitude mean value, an amplitude variance, a peak-to-peak value, a waveform factor, a kurtosis factor, a pulse factor, a permutation entropy, a singularity entropy, a sample entropy, a wavelet packet energy, a 4-layer wavelet packet singular value and a 6-layer wavelet packet singular value of the driving torque signal of the hydraulic cylinder.

[0019] Preferably, the deep belief network model includes an input layer, three hidden layers and an output layer;

[0020] wherein the number of nodes of the input layer corresponds to the dimension number of the feature vector, the number of nodes of the three hidden layers decreases successively, and the number of nodes of the output layer corresponds to the number of division levels of the coal rock hardness.

[0021] Preferably, the number of nodes of the three hidden layers is 36, 18 and 9 successively.

[0022] Preferably, the cutting motion parameter database is established by the following steps:

[0023] Step S411, a comprehensive cutting performance multi-objective optimization model is constructed, the design variables of which are the swing speed of the cutting arm and the rotating speed of the cutting head, and the objective function of which is a weighted sum function constructed based on the coal rock production rate, the average cutting area of the pick, the cutting specific energy consumption and the gear dynamic load;

[0024] Step S412, the multi-objective optimization model is iteratively solved by using the pre-constructed improved whale optimization algorithm to obtain the parameter combination of the recommended rotating speed of the cutting head and the recommended swing speed of the cutting arm that optimizes the objective function under different coal rock hardness;

[0025] Step S413, the parameter combination is stored in association with the corresponding coal rock hardness to form the cutting motion parameter database.

[0026] Preferably, the improved whale optimization algorithm performs the following operations in each iteration process:

[0027] The global optimal solution position in the last iteration period is recorded and reserved;

[0028] The distance between each individual in the current population and the current global optimal solution position is calculated, and each individual in the population is sequentially assigned an increasing serial number and a sequence table is generated according to the numerical size of the distance;

[0029] For the individuals with the last 50% of the serial numbers in the sequence table, the position is updated according to the randomly selected individual position;

[0030] For the individuals with the first 50% of the serial numbers in the sequence table, the spiral position is updated according to the global optimal solution position in the last iteration period.

[0031] Preferably, in the weighted sum function, the absolute values of the weight coefficients of the coal rock production rate, the average cutting area of the pick, the cutting specific energy consumption and the dynamic load of the transmission system are respectively set to 0.4, 0.1, 0.3 and 0.2.

[0032] Preferably, in the step S500, the rotating speed of the cutting head and the swing speed of the cutting arm are configured to have the same dynamic response time when they are controlled cooperatively.

[0033] Preferably, when the hardness increases, the rotating speed of the cutting head is controlled to smoothly decrease along a concave function trajectory; when the hardness decreases, the rotating speed of the cutting head is controlled to smoothly increase along a linear slope trajectory.

[0034] An intelligent cutting control system of a heading machine, which is applied in the intelligent cutting control method of the heading machine described above, comprises;

[0035] The signal acquisition module is used to acquire feedback signals of the tunneling machine in real time during the cutting process through sensors. The feedback signals include the current signal of the cutting motor, the vibration acceleration signal of the cutting arm, and the driving torque signal of the hydraulic cylinder.

[0036] The feature extraction module is used to calculate multi-domain feature parameters in the time domain, frequency domain, and entropy value of the acquired feedback signal to generate feature vectors;

[0037] The hardness recognition module is used to input the feature vector into a pre-constructed deep belief network model for hardness recognition and output the hardness of the currently cut coal and rock.

[0038] The parameter matching module is used to retrieve a preset cutting motion parameter database based on the hardness and obtain a parameter combination corresponding to the current cutting coal and rock hardness. The parameter combination includes the recommended rotation speed of the cutting head and the recommended swing speed of the cutting arm.

[0039] The collaborative control module, based on the acquired parameter combinations, is used to collaboratively control the cutting head speed and cutting arm swing speed of the tunneling machine.

[0040] The beneficial effects of this invention are:

[0041] This invention provides an intelligent cutting control method and system for tunneling machines. Through multi-domain feature extraction, deep belief network intelligent recognition, improved whale optimization algorithm to construct a parameter database, and coordinated control of the cutting head rotation speed and cutting arm swing speed, the cutting process of the tunneling machine is comprehensively optimized. Therefore, the system provided by this invention can accurately sense changes in coal and rock hardness and automatically adjust the cutting parameters to the optimal state of comprehensive performance, thereby effectively improving cutting efficiency, reducing energy consumption, reducing equipment wear, and enhancing the adaptability and working stability under complex coal seam conditions. Attached Figure Description

[0042] Figure 1 A schematic diagram of an intelligent cutting control system for a tunneling machine provided in one embodiment of the present invention;

[0043] Figure 2 This is a schematic diagram of the structure of a tunneling machine provided in one embodiment of the present invention.

[0044] In the diagram, 1. Vibration acceleration sensor; 2. Cutting arm; 3. Cutting motor; 4. Pitch angle encoder; 5. Turntable; 6. Slewing angle encoder; 7. Slewing bearing; 8. Hydraulic pressure sensor; 9. Tunneling machine chassis; 10. Cutting motor driver; 11. Industrial control computer; 12. Signal conditioning box; 13. Electro-hydraulic directional valve. Detailed Implementation

[0045] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0046] Example 1, referring to Figure 1 This invention provides an intelligent cutting control method for tunneling machines, which includes the following steps:

[0047] Step S100:

[0048] During the cutting operation of the tunneling machine, the system simultaneously collects three types of feedback signals: the current signal of the cutting motor, the vibration acceleration signal of the cutting arm, and the driving torque signal of the hydraulic cylinder. The cutting motor current signal reflects the load on the cutting head; the current increases significantly when cutting hard rock and is relatively lower when cutting soft coal. The cutting arm vibration acceleration signal contains rich information about the cutting state; the vibration frequency components and amplitude characteristics generated when cutting coal and rock of different hardnesses differ significantly. The hydraulic cylinder driving torque is calculated by converting the cylinder output force and the cutting head position, directly reflecting the resistance torque during the swinging process of the cutting arm.

[0049] In practical implementation, the present invention provides a preset sliding time window to continuously collect the feedback signals from the sensor within the last 3 seconds, ensuring the real-time nature and continuity of the data.

[0050] Step S200:

[0051] This invention performs in-depth processing on the acquired raw feedback signal, extracting 36 feature parameters from three dimensions: time domain, frequency domain, and nonlinear dynamics, to construct a feature vector that comprehensively describes the cut-off state.

[0052] The extraction of time-domain features in this invention includes conventional statistics and dimensionless indices. Conventional statistics calculate the mean amplitude, variance, and peak-to-peak value of current, torque, and vibration signals. The mean amplitude reflects the static component of the signal, the variance characterizes the degree of signal fluctuation, and the peak-to-peak value indicates extreme cases. Dimensionless indices include waveform factor, kurtosis factor, and impulse factor. The waveform factor is the ratio of the effective value to the absolute mean, reflecting waveform characteristics; the kurtosis factor characterizes the sharpness of the signal distribution and is sensitive to impact components; the impulse factor is the ratio of the peak value to the absolute mean and has a good indicative effect on instantaneous impacts.

[0053] In this invention, the extraction of frequency domain features primarily targets vibration acceleration signals. A Fast Fourier Transform (FFT) is used to convert the time-domain signal into a frequency-domain representation. The centroid frequency represents the center of energy distribution in the spectrum, and its calculation formula is the weighted average of the amplitudes of each frequency component. The mean square frequency reflects the dispersion of the frequency distribution, and the variance frequency characterizes the dispersion of the spectrum. Through these frequency domain features, this invention can effectively distinguish the differences in vibration frequency characteristics generated during the cutting of coal and rock of different hardness.

[0054] The entropy multi-domain feature in this invention includes entropy analysis and wavelet packet decomposition.

[0055] Among them, entropy analysis calculates the permutation entropy, singular entropy, and sample entropy of a signal, which are used to quantify the complexity and randomness of the signal. Permutation entropy assesses complexity by examining the ordering patterns of adjacent values ​​in a time series; singular entropy, based on singular spectrum analysis, reflects the complexity of signal dynamics; and sample entropy measures the probability of a time series generating new patterns and is sensitive to nonlinear characteristics.

[0056] Wavelet packet decomposition employs the db4 wavelet basis function for 4-level and 6-level decomposition to extract wavelet packet energy features from each node. The 4-level decomposition generates 16 frequency bands, and the 6-level decomposition generates 64 frequency bands. By calculating the proportion of energy in each frequency band to the total energy, a refined frequency band energy distribution characteristic is obtained. Therefore, in this invention, wavelet packet singular values ​​are obtained through singular value decomposition of the signals in each frequency band, reflecting the main energy component characteristics of the signal.

[0057] Specifically, this invention extracts 12 features from the cutting motor current signal: mean amplitude, variance, peak-to-peak value, waveform factor, kurtosis factor, impulse factor, permutation entropy, singular entropy, sample entropy, wavelet packet energy, 4-layer wavelet packet singular values, and 6-layer wavelet packet singular values. Similarly, the same 12 features are extracted from the hydraulic drive torque signal. The cutting arm vibration acceleration signal extracts 12 features: mean amplitude, variance, kurtosis, center of gravity frequency, mean square frequency, variance frequency, permutation entropy, singular entropy, sample entropy, wavelet packet energy, 4-layer wavelet packet singular values, and 6-layer wavelet packet singular values. Finally, this invention outputs a 36-dimensional feature vector, providing a sufficient data foundation for subsequent hardness identification.

[0058] Step S300:

[0059] The deep belief network model used in this invention has a unique deep structure, including an input layer, three hidden layers, and an output layer. The input layer has 36 nodes, consistent with the dimension of the feature vector; the three hidden layers have 36, 18, and 9 nodes respectively, forming a funnel-shaped structure; in this embodiment, the output layer has 7 nodes, corresponding to the preset 7 levels of coal and rock hardness (0-15MPa soft coal, 15-26MPa medium-hard coal, 26-37MPa transition between medium-hard and hard coal, 37-48MPa hard coal and extremely hard coal, 48-58MPa extremely hard coal and interbedded coal, 58-69MPa interbedded coal and relatively firm rock, 69-80MPa firm rock).

[0060] In this invention, the training of the DBN model is divided into two stages: unsupervised pre-training and supervised fine-tuning. In the pre-training stage, multiple stacked Restricted Boltzmann Machines (RBMs) are trained. The visible layer of the first RBM receives a 36-dimensional feature vector, while the hidden layer of the previous RBM serves as the visible layer of the next RBM, and so on, stacking layer by layer. Through this process, the invention learns a hierarchical feature representation of the input data and initializes the network weights to a better state. The fine-tuning stage employs the backpropagation algorithm, using labeled sample data to globally optimize the network parameters and minimize prediction error.

[0061] In this embodiment, the training data used in this invention comes from a large number of coal and rock cutting experiments, with 800 sets of data collected for each coal and rock sample of each hardness level. Specifically, in this embodiment, 50% of the dataset is used to train the DBN recognition model, and 50% is used to test and verify the generalization ability of the trained model. When training the DBN globally, the learning rate is set to 0.1, the momentum factor is set to 0.9, and the number of iterations is 200.

[0062] In practical applications, the trained DBN model operates in a forward propagation manner: the input is a 36-dimensional feature vector, which undergoes nonlinear transformation through three hidden layers, and finally the output layer obtains the probability of belonging to 7 hardness levels. The level corresponding to the highest probability is taken as the hardness recognition result.

[0063] Step S400:

[0064] The cutting motion parameter database is a core component of this invention, and its construction process is based on the improved whale optimization algorithm and the multi-objective optimization model for comprehensive cutting performance involved in this invention. The cutting motion parameter database is stored in the form of a lookup table, where the key is the coal / rock hardness level and the value is the corresponding optimal combination of cutting parameters (cutting head rotation speed and cutting arm swing speed).

[0065] The comprehensive multi-objective optimization model for cutting performance considers four key performance indicators: coal and rock productivity, average cutting area of ​​the cutting teeth, specific energy consumption for cutting, and gear dynamic load. The objective function for coal and rock productivity is represented by the volume of coal and rock cut per unit time, which is positively correlated with the swing speed of the cutting arm but independent of the cutting head speed. The average cutting area of ​​the cutting teeth affects the wear rate of the cutting teeth; an excessively small cutting area will increase dust on the working face and exacerbate the wear and consumption of the cutting teeth. Specific energy consumption for cutting is defined as the energy consumed in cutting a unit volume of coal and rock, and is an important indicator for evaluating the economic efficiency of cutting. The gear dynamic load is represented by the instantaneous change in the meshing force between the sun gear and planetary gears of the planetary gear when the hardness of the coal and rock changes by 1.5 times, reflecting the impact of external impact loads on the operating performance of the transmission system.

[0066] The description of the multi-objective optimization model for comprehensive cut-off performance is as follows:

[0067]

[0068] in, for transpose matrix, ~ These are the weighting coefficients for coal and rock productivity, cutting area of ​​the cutting teeth, specific energy consumption for cutting, and dynamic load of the gears. Combined with the actual production needs of coal mining enterprises, [the following are considered]... ~ The values ​​were set to 0.4, 0.1, 0.3, and 0.2 respectively. This means that when optimizing the cutting parameters, the impact of productivity and cutting ratio energy consumption should be given priority to improve the cutting efficiency of the tunneling machine. Secondly, the dynamic load should be reduced to improve the reliability of the cutting transmission system and suppress coal and rock cutting dust.

[0069] Meanwhile, considering the significant differences in the dimensions and orders of magnitude of different indicators, the four cutoff performance indicators are normalized. Let the first... ( The range of the sub-objective functions (=1~4) is... The transformation is performed using the min-max normalization function. That is, when At that time, the sub-target is at its maximum value. ,when When the target is at its minimum value, .

[0070] Specifically, the sub-objectives in the comprehensive cutting performance multi-objective optimization model mainly consider key performance indicators such as coal and rock productivity, cutting area of ​​the cutting teeth, cutting specific energy consumption, and gear dynamic load. The mathematical models of the sub-objectives of the relevant cutting performance indicators are introduced as follows:

[0071] (1) Coal and rock productivity

[0072] The sub-objective model corresponding to the productivity of tunneling machine cutting coal and rock is represented as follows:

[0073] ;

[0074] in, This is the coal and rock loosening coefficient, which can generally be taken as 1.5. This refers to the axial cross-sectional area of ​​the cutting head. The cutting head swing speed.

[0075] (2) Average cutting area of ​​the cutting teeth

[0076] The sub-objective model corresponding to the average cutting area of ​​the main cutting section teeth of the tunneling machine cutter head is as follows:

[0077] ;

[0078] In the formula, For cutting teeth The corresponding cut-off groove is projected onto the striking surface.

[0079] (3) Cutting energy consumption

[0080] Cutting energy consumption reflects the energy utilization rate of a tunneling machine during cutting and is one of the important economic indicators for measuring the working performance of a tunneling machine. The corresponding sub-objective model is expressed as:

[0081] ;

[0082] in, For coal and rock cutting torque, Indicates the angular velocity of the cutting head. For coal and rock cutting force, This refers to coal and rock productivity.

[0083] (4) Sudden dynamic load on gears

[0084] The instantaneous change in meshing force between the sun gear and planet gears of the first stage of the planetary gear when the coal and rock hardness changes by 1.5 times is taken as the sub-objective for optimizing the dynamic load of the lower gear. The sub-objective model corresponding to the gear dynamic load is expressed as follows:

[0085] ;

[0086] In the formula, This indicates the peak value of the dynamic load during gear mutation. This indicates the rated meshing force of the gear pair.

[0087] Furthermore, the implementation of the improved whale optimization algorithm includes the following steps:

[0088] First, initialize the population size. =30, maximum number of iterations =100, and the upper and lower bounds of the parameter are determined according to the on-site working requirements of the tunneling machine. In each iteration, the fitness value (i.e., the multi-objective function value) of each individual whale is calculated, and then the improved position update strategy is executed.

[0089] The core of the improved strategy lies in two points: first, using the historical best individual instead of the current best individual to avoid getting trapped in local optima; second, sorting the population according to its distance from the best individual, performing random search updates on the bottom 50% of individuals, and performing spiral position updates on the top 50% of individuals toward the historical best individual.

[0090] Therefore, the formula for updating the specific location is:

[0091] For the bottom 50% of individuals who are further away:

[0092] ;

[0093] ;

[0094] in, For individual whales The position of the wheel; This indicates the location of a randomly selected individual whale within the group;

[0095] For the top 50% of individuals who are closest to each other:

[0096] ;

[0097] ;

[0098] ;

[0099] in, This indicates the position of the optimal solution in the current group. Represents a random number between [0, 2]. Represents the helix constant; Represents a random number in the interval [-1, 1]. This indicates the initialization parameters.

[0100] Through iterative optimization, the optimal combination of cutting parameters for each hardness level was obtained, forming a cutting motion parameter database as shown in Table 1:

[0101] Table 1: Example of a database of cutting motion parameters

[0102] Therefore, in the actual control process, the system directly queries the table based on the hardness level identified by DBN to obtain the corresponding recommended cutting head speed and recommended cutting arm swing speed, which are used as the set values ​​for the control system.

[0103] Step S500:

[0104] The collaborative control module receives the recommended rotational speed and swing speed from the parameter matching module, and uses advanced control algorithms to coordinate the movements of the cutting head and cutting arm. The core requirement of the control system is to ensure that the rotational speed and swing speed simultaneously reach the set values, and that the dynamic adjustment process is smooth and stable.

[0105] The cutting head speed control employs a direct torque control strategy. The system monitors the cutting motor torque and flux linkage in real time, compares the measured values ​​with the setpoint using a hysteresis comparator, and directly generates the inverter's switching signal based on the error status by querying the switching table. The torque setpoint is generated by a PID controller based on the recommended speed, and the speed controller output serves as the reference input for the inner torque loop. This dual closed-loop structure used in this invention ensures both the accuracy of speed control and rapid torque response.

[0106] The swing speed control of the cutting arm employs a backstepping sliding mode cascade control algorithm. First, a nonlinear mathematical model of the cutting arm's hydraulic drive system is established, decomposing the system into a position subsystem and a pressure subsystem. Based on Lyapunov stability theory, virtual control variables and the final control law are designed step-by-step. The sliding surface is designed as a linear combination of system state errors, and by selecting an appropriate switching gain, the system state is ensured to converge to the sliding surface within a finite time. To mitigate the inherent chattering problem of sliding mode control, the tanh function is used instead of the sign function.

[0107] The key to the collaborative mechanism lies in the consistent configuration of dynamic response time. Even if the cutting head speed and the cutting arm swing speed have the same control time, the adjustment time is usually set to 0.5 to 1.0 seconds. This ensures that when the hardness of the coal and rock changes, the cutting speed of the cutting head and the feed speed of the cutting arm are adjusted synchronously, avoiding non-stationary changes in the system caused by inconsistent actions of the two.

[0108] The trajectory planning is specifically designed for situations involving sudden changes in coal and rock hardness. When hardness increases, the cutting head speed decreases along a parabolic trajectory. This trajectory changes rapidly in the initial stage, which helps the cutting motor to unload quickly. Conversely, when hardness decreases, the speed increases along a ramp trajectory, thereby achieving smooth acceleration. The adjustment time is adaptively adjusted according to the magnitude of the hardness change; the greater the change, the longer the adjustment time, ensuring a smooth adjustment process.

[0109] Example 2: The present invention provides an intelligent cutting control system for a tunneling machine, which is applied to the above-mentioned intelligent cutting control method for a tunneling machine, including:

[0110] The signal acquisition module is used to acquire feedback signals of the tunneling machine in real time during the cutting process through sensors. The feedback signals include the current signal of the cutting motor, the vibration acceleration signal of the cutting arm, and the driving torque signal of the hydraulic cylinder.

[0111] The feature extraction module is used to calculate multi-domain feature parameters in the time domain, frequency domain, and entropy value of the acquired feedback signal to generate feature vectors;

[0112] The hardness recognition module is used to input the feature vector into a pre-constructed deep belief network model for hardness recognition and output the hardness of the currently cut coal and rock.

[0113] The parameter matching module is used to retrieve a preset cutting motion parameter database based on the hardness and obtain a parameter combination corresponding to the current cutting coal and rock hardness. The parameter combination includes the recommended rotation speed of the cutting head and the recommended swing speed of the cutting arm.

[0114] The collaborative control module, based on the acquired parameter combinations, is used to collaboratively control the cutting head speed and cutting arm swing speed of the tunneling machine.

[0115] Reference Figure 2 The current signal of the cutting motor is measured by the industrial general-purpose motor driver 10. The motor driver 10 is connected to a 380V three-phase power supply. Its input port is connected to the control board port of the industrial control computer 11 to receive the speed control signal. Its output port is connected to the cutting motor 3. Both the driver 10 and the industrial control computer 11 are installed in the electrical control cabinet through vibration isolation plates. The vibration acceleration signal of the cutting arm is measured by the piezoelectric acceleration sensor 1. The acceleration sensor 1 is installed on the cutting arm 2 near the cutting head. Two absolute encoders 4 and 6 are installed at the rotation shafts of the cutting arm 2 and the rotary table 5, and the rotary table 5 and the slewing bearing 7 to measure the pitch / horizontal swing angle of the cutting head. The slewing bearing 7 is rigidly connected to the tunneling machine chassis 9. In order to measure the hydraulic pressure of the two chambers of the pitch cylinder and the slewing cylinder, four thin-film pressure sensors 8 are installed on the connecting pipeline between the cylinder and the hydraulic control valve. The driving torque signal of the hydraulic cylinder is calculated from the collected swing angle signal and pressure signal.

[0116] Specifically, the sensors configured in the signal acquisition module are connected to the signal conditioning box 12. The signal conditioning box 12 converts and processes the acquired signals and sends them to the acquisition board of the industrial control computer 11. The hardness identification module calculates multi-domain feature parameters in the time domain, frequency domain, and entropy value through a pre-programmed feature extraction module, and then determines the approximate hardness range of the coal and rock. Based on the pre-established cutting motion parameter database, it outputs the optimal combination of cutting head speed and cutting arm swing speed parameters to the collaborative control module. The collaborative control module performs online planning based on the current cutting head speed and cutting arm swing speed and their reference values, and dynamically corrects the error between the expected value and the feedback value. The control board of the industrial control computer 11 performs servo control on the cutting motor speed and the cutting arm swing speed. The industrial control computer 11 sends speed commands to the motor driver 10 to adjust the cutting head speed. At the same time, the industrial control computer sends flow commands to the electro-hydraulic directional valve 13 to dynamically adjust the cutting arm swing speed.

[0117] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0118] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for intelligent cutting control of a tunneling machine, characterized in that, Includes the following steps: Step S100: The feedback signals of the tunneling machine during the cutting process are collected in real time through sensors; the feedback signals include the current signal of the cutting motor, the vibration acceleration signal of the cutting arm, and the driving torque signal of the hydraulic cylinder. Step S200: Calculate the multi-domain feature parameters (time domain, frequency domain, and entropy value) of the collected feedback signal to generate a feature vector; Step S300: Input the feature vector into the pre-constructed deep belief network model for hardness identification and output the hardness of the currently cut coal and rock. Step S400: Based on the hardness, retrieve the preset cutting motion parameter database to obtain the parameter combination corresponding to the current cutting coal and rock hardness; wherein, the parameter combination includes the recommended rotation speed of the cutting head and the recommended swing speed of the cutting arm; Step S500: Based on the acquired parameter combination, the cutting head speed and cutting arm swing speed of the tunneling machine are controlled in a coordinated manner.

2. The intelligent cutting control method for tunneling machines as described in claim 1, characterized in that, The calculation of multi-domain characteristic parameters (time domain, frequency domain, and entropy value) of the acquired feedback signal includes: Calculate the amplitude mean, amplitude variance, peak-to-peak value, waveform factor, kurtosis factor, impulse factor, permutation entropy, singular entropy, sample entropy, wavelet packet energy, 4-layer wavelet packet singular values, and 6-layer wavelet packet singular values ​​of the cut motor current signal; Calculate the mean amplitude, variance amplitude, kurtosis amplitude, centroid frequency, mean square frequency, variance frequency, permutation entropy, singular entropy, sample entropy, wavelet packet energy, 4-layer wavelet packet singular values, and 6-layer wavelet packet singular values ​​of the vibration acceleration signal of the cutting arm; calculate the mean amplitude, variance amplitude, peak-to-peak value, waveform factor, kurtosis factor, impulse factor, permutation entropy, singular entropy, sample entropy, wavelet packet energy, 4-layer wavelet packet singular values, and 6-layer wavelet packet singular values ​​of the hydraulic cylinder drive torque signal.

3. The intelligent cutting control method for tunneling machines as described in claim 2, characterized in that, The deep belief network model comprises an input layer, three hidden layers, and an output layer; The number of nodes in the input layer corresponds to the dimension of the feature vector, the number of nodes in the three hidden layers decreases sequentially, and the number of nodes in the output layer corresponds to the number of coal and rock hardness classification levels.

4. The intelligent cutting control method for tunneling machines as described in claim 3, characterized in that, The number of nodes in the three hidden layers are 36, 18, and 9, respectively.

5. The intelligent cutting control method for tunneling machines as described in claim 1, characterized in that, The cutting motion parameter database is established through the following steps: Step S411: Construct a multi-objective optimization model for comprehensive cutting performance. The design variables are the swing speed of the cutting arm and the rotation speed of the cutting head. The objective function is a weighted sum function based on coal and rock productivity, average cutting area of ​​the cutting teeth, cutting specific energy consumption and gear dynamic load. Step S412: The pre-constructed improved whale optimization algorithm is used to iteratively solve the multi-objective optimization model to obtain the parameter combination of recommended cutting head rotation speed and recommended cutting arm swing speed that optimizes the objective function under different coal and rock hardness. Step S413: The parameter combination is associated with the corresponding coal and rock hardness and stored to form the cutting motion parameter database.

6. The intelligent cutting control method for tunneling machines as described in claim 5, characterized in that, The improved whale optimization algorithm performs the following operations in each iteration: Record and retain the position of the global optimal solution from the previous iteration cycle; Calculate the distance between each individual in the current population and the current global optimal solution, sort each individual in the population in descending order according to the value of the distance, assign them an increasing sequence number, and generate a sequence list. For the last 50% of individuals in the sequence list, the position is updated based on the randomly selected individual position; For the individuals in the first 50% of the sequence numbers in the sequence list, a spiral position update is performed based on the position of the global optimal solution in the previous iteration cycle.

7. The intelligent cutting control method for tunneling machines as described in claim 5, characterized in that, In the weighted sum function, the absolute values ​​of the weighting coefficients for coal and rock productivity, average cutting area of ​​the cutting teeth, cutting ratio energy consumption, and dynamic load of the transmission system are set to 0.4, 0.1, 0.3, and 0.2, respectively.

8. The intelligent cutting control method for tunneling machines as described in claim 1, characterized in that, In step S500, when the cutting head rotation speed and the cutting arm swing speed are controlled in a coordinated manner, they are configured to have the same dynamic response time.

9. The intelligent cutting control method for tunneling machines as described in claim 8, characterized in that, When the hardness increases, the rotational speed of the cutting head is controlled to decrease smoothly along a concave function trajectory; when the hardness decreases, the rotational speed of the cutting head is controlled to increase smoothly along a linear ramp trajectory.

10. A tunneling machine intelligent cutting control system, applied in the tunneling machine intelligent cutting control method as described in any one of claims 1-9, characterized in that, include; The signal acquisition module is used to acquire feedback signals of the tunneling machine in real time during the cutting process through sensors. The feedback signals include the current signal of the cutting motor, the vibration acceleration signal of the cutting arm, and the driving torque signal of the hydraulic cylinder. The feature extraction module is used to calculate multi-domain feature parameters in the time domain, frequency domain, and entropy value of the acquired feedback signal to generate feature vectors; The hardness recognition module is used to input the feature vector into a pre-constructed deep belief network model for hardness recognition and output the hardness of the currently cut coal and rock. The parameter matching module is used to retrieve a preset cutting motion parameter database based on the hardness and obtain a parameter combination corresponding to the current cutting coal and rock hardness. The parameter combination includes the recommended rotation speed of the cutting head and the recommended swing speed of the cutting arm. The collaborative control module, based on the acquired parameter combinations, is used to collaboratively control the cutting head speed and cutting arm swing speed of the tunneling machine.

Images