Early warning method and device for abnormal state of hobbing cutter of heading machine

Abstract

The invention discloses a heading machine hob abnormal state early warning method and device. The method comprises the steps that real-time heading parameter data, hob vibration data, hob infrared video data, hob temperature data and rock slag image data are acquired; acquiring historical tunneling parameter data according to a specified mileage, inputting the historical tunneling parameter data and the real-time tunneling parameter data into a first hob damage early warning model, and outputting a first early warning result; inputting the hob vibration data into a second hob damage early warning model, and outputting a second early warning result; inputting the hob infrared video data and the hob temperature data into a third hob damage early warning model, and outputting a third early warning result; inputting the rock slag image data into a fourth hob damage early warning model, and outputting a fourth early warning result; and based on the Bayesian theorem, fusing the first early warning result, the second early warning result, the third early warning result and the fourth early warning result to obtain a hob state early warning result. The accuracy of early warning of the abnormal state of the hob can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tunnel construction, in particular to a tunneling machine cutterhead abnormal state early warning method and device. BACKGROUND

[0002] The cutterhead is one of the key components of the tunneling machine, and during the construction process of the tunneling machine, the cutterhead will have different degrees of wear, uneven wear and chipping, which seriously affects the tunneling efficiency. The site personnel need to frequently stop the machine and open the bin to check the wear state of the cutter, and if necessary, replace the cutter, which increases the construction cost.

[0003] The existing cutter state early warning methods include mechanical analysis methods and artificial intelligence analysis methods. The mechanical analysis method uses the mechanical relationship between the cutter and the rock mass to perform mechanical analysis on the basis of the tunneling parameters detected on site, thereby predicting the wear of the cutter in real time. However, in this method, the physical model is generally simplified and many restrictions are set, which makes it impossible to accurately predict the state of the cutter. The artificial intelligence analysis method obtains the tunneling parameters or indirectly obtains the cutter wear information through detection devices for analysis and processing. However, the existing artificial intelligence analysis method has poor robustness and is easily affected by data changes, and the prediction accuracy needs to be improved. SUMMARY

[0004] The tunneling machine cutterhead abnormal state early warning method provided by the embodiments of the present application can analyze the cutter damage factors from multiple scales and multiple directions, monitor and warn the cutter state in all directions, and improve the accuracy of the cutter abnormal state early warning. The method comprises the following steps:

[0005] Obtain real-time tunneling parameter data, cutter vibration data, cutter infrared video data, cutter temperature data, and rock slag image data.

[0006] Obtain historical tunneling parameter data at a specified mileage, input the historical tunneling parameter data and real-time tunneling parameter data into a first cutter damage early warning model, and output a first early warning result. The first cutter damage early warning model is used to compare and analyze the trends of the historical tunneling parameter data, real-time tunneling parameter data and historical tunneling parameters when the cutter is abnormal, and determine the first early warning result according to the trend comparison and analysis result.

[0007] Input the cutter vibration data into a second cutter damage early warning model, and output a second early warning result. The second cutter damage early warning model is trained by using historical cutter vibration data and cutter damage state data in a gated recurrent neural network.

[0008] The cutter infrared video data and the cutter temperature data are input into a third cutter damage early warning model, and a third early warning result is output; the third cutter damage early warning model is used for: determining the third early warning result by using a pre-established mapping relationship between a cutter damage state and a temperature value during cutter operation, the cutter infrared video data and the cutter temperature data;

[0009] The rock slag image data are input into a fourth cutter damage early warning model, and a fourth early warning result is output; the fourth cutter damage early warning model is used for: determining a current rock slag characteristic value according to the rock slag image data, and determining the fourth early warning result by using a pre-established mapping relationship between a cutter damage state and a rock slag characteristic during cutter operation and the current rock slag characteristic value;

[0010] Based on the Bayes theorem, the first early warning result, the second early warning result, the third early warning result and the fourth early warning result are fused to obtain a cutter state early warning result.

[0011] The embodiment of the application also provides a tunneling machine cutter abnormal state early warning device for multi-scale and multi-directional analysis of cutter damage factors, omnidirectional monitoring and early warning of a cutter state and improvement of the accuracy of cutter abnormal state early warning, the device comprising:

[0012] A data acquisition module is used for acquiring real-time tunneling parameter data, cutter vibration data, cutter infrared video data, cutter temperature data and rock slag image data.

[0013] A first sub-model processing module is used for acquiring historical tunneling parameter data according to a specified mileage, inputting the historical tunneling parameter data and the real-time tunneling parameter data into a first cutter damage early warning model and outputting a first early warning result; the first cutter damage early warning model is used for: performing trend comparison and analysis on the historical tunneling parameter data, the real-time tunneling parameter data and historical tunneling parameters during cutter abnormality, and determining the first early warning result according to a trend comparison and analysis result;

[0014] A second sub-model processing module is used for inputting the cutter vibration data into a second cutter damage early warning model and outputting a second early warning result; the second cutter damage early warning model is pre-trained by using historical cutter vibration data and cutter damage state data;

[0015] A third sub-model processing module is used for inputting the cutter infrared video data and the cutter temperature data into a third cutter damage early warning model and outputting a third early warning result; the third cutter damage early warning model is used for: determining the third early warning result by using a pre-established mapping relationship between a cutter damage state and a temperature value during cutter operation, the cutter infrared video data and the cutter temperature data;

[0016] The fourth sub-model processing module is used to input rock slag image data into the fourth cutter damage early warning model and output the fourth early warning result; the fourth cutter damage early warning model is used to: determine the current rock slag feature value based on the rock slag image data, and determine the fourth early warning result using the pre-established mapping relationship between the cutter damage state and rock slag features during cutter operation and the current rock slag feature value;

[0017] The fusion processing module is used to fuse the first, second, third, and fourth warning results based on Bayes' theorem to obtain the rolling cutter status warning result.

[0018] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described method for early warning of abnormal conditions of the tunneling machine cutter head.

[0019] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for early warning of abnormal conditions of the tunneling machine cutter head.

[0020] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described method for early warning of abnormal conditions of the tunneling machine cutter head.

[0021] In this embodiment of the invention, real-time tunneling parameter data, cutter vibration data, cutter infrared video data, cutter temperature data, and rock debris image data are acquired, and four sub-models are constructed. Each sub-model is used to process and analyze the real-time tunneling parameter data, cutter vibration data, cutter infrared video data, cutter temperature data, and rock debris image data. Finally, the warning results of the four sub-models are combined for fusion judgment. This achieves comprehensive consideration of multi-scale and multi-directional cutter damage factors, avoiding the problem of warning failure caused by the failure of a single warning device. Single data corruption or sudden changes will not have a significant impact on the warning results, greatly improving the accuracy of cutter abnormality warnings and realizing comprehensive monitoring and warning of cutter status. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0023] Figure 1This is a flowchart illustrating the early warning method for abnormal conditions of the tunneling machine cutter head in an embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram of a multi-source data acquisition system in an embodiment of the present invention;

[0025] Figure 3 This is a schematic diagram of the tunneling parameter data feature analysis in an embodiment of the present invention;

[0026] Figure 4 This is a schematic diagram of vibration data feature analysis in an embodiment of the present invention;

[0027] Figure 5 This is a schematic diagram illustrating the relationship between the damage state of the hob and temperature data in an embodiment of the present invention;

[0028] Figure 6 This is a schematic diagram of slag image analysis in an embodiment of the present invention. Figure 1 ;

[0029] Figure 7 This is a schematic diagram of slag image analysis in an embodiment of the present invention. Figure 2 ;

[0030] Figure 8 This is a schematic diagram of an early warning device for abnormal status of the tunneling machine cutter head in an embodiment of the present invention. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0032] The acquisition, storage, use, and processing of data in this application all comply with the relevant provisions of national laws and regulations.

[0033] Existing cutter condition warning systems lack accuracy. Therefore, this invention proposes a method for early warning of abnormal cutter conditions in tunneling machines. This method analyzes cutter damage factors from multiple scales and perspectives, providing comprehensive monitoring of the cutter condition and enabling intelligent diagnosis and accurate real-time early warning of abnormal cutter conditions. This provides tunneling machine operators with accurate cutter warning information, reducing safety accidents caused by severe cutter damage, making cutter replacement more efficient, thereby reducing construction costs and improving tunneling efficiency.

[0034] Figure 1 This is a flowchart illustrating the early warning method for abnormal conditions of the tunneling machine cutter head in an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes:

[0035] Step 101: Acquire real-time tunneling parameter data, cutter vibration data, cutter infrared video data, cutter temperature data, and rock debris image data;

[0036] Step 102: Obtain historical tunneling parameter data according to the specified mileage, input the historical tunneling parameter data and real-time tunneling parameter data into the first cutter head damage early warning model, and output the first early warning result; the first cutter head damage early warning model is used to: perform trend comparison analysis on the historical tunneling parameter data, real-time tunneling parameter data and historical tunneling parameters when the cutter head is abnormal, and determine the first early warning result based on the trend comparison analysis result;

[0037] Step 103: Input the hob vibration data into the second hob damage early warning model and output the second early warning result; the second hob damage early warning model is obtained by training a gated recurrent neural network in advance using historical hob vibration data and hob damage status data.

[0038] Step 104: Input the cutter infrared video data and cutter temperature data into the third cutter damage early warning model and output the third early warning result; the third cutter damage early warning model is used to: determine the third early warning result by using the pre-established mapping relationship between the cutter damage state and temperature value during cutter operation, the cutter infrared video data and the cutter temperature data.

[0039] Step 105: Input the rock slag image data into the fourth cutter damage early warning model and output the fourth early warning result; the fourth cutter damage early warning model is used to: determine the current rock slag feature value based on the rock slag image data, and determine the fourth early warning result using the pre-established mapping relationship between the cutter damage state and rock slag features during cutter operation and the current rock slag feature value.

[0040] Step 106: Based on Bayes' theorem, the first warning result, the second warning result, the third warning result and the fourth warning result are merged to obtain the rolling cutter state warning result.

[0041] The following is a detailed explanation of the early warning method for abnormal status of the tunneling machine cutter in the embodiments of the present invention.

[0042] In step 101, real-time tunneling parameter data, cutter vibration data, cutter infrared video data, and rock debris image data are acquired.

[0043] During the tunneling process, the cutter head will experience varying degrees of uneven wear or even chipping. The most obvious characteristic of the cutter head cutting rock is the generation of vibration and heat. Abnormal uneven wear or chipping of the cutter head will inevitably cause abnormal vibration of the cutter head and generate high temperature.

[0044] Vibration sensors are installed on the cutterhead to acquire cutterhead vibration data, and the relationship between vibration and cutterhead damage is analyzed. Infrared sensors are installed behind the cutterhead to acquire infrared video stream data and cutterhead temperature data during cutterhead rotation, and abnormal temperature characteristics when cutterhead damage occurs are extracted. When cutterhead damage occurs in a localized area on the cutterhead, tunneling parameters such as cutterhead speed, cutterhead torque, and thrust will show corresponding abnormal changes. Simultaneously, the integrity and hardness of the surrounding rock also affect the frequency of cutterhead damage when the tunneling machine is tunneling through different geological layers. Therefore, in this embodiment of the invention, the above-mentioned multimodal data is acquired, and then, based on multimodal fusion technology, the data of different modes are effectively processed separately, and the decision results of each are fused to obtain the final early warning decision. This provides on-site personnel with multi-scale and more accurate early warning information on cutterhead damage, providing an effective basis for on-site cutterhead replacement.

[0045] In one embodiment, acquiring real-time tunneling parameter data, cutter vibration data, cutter infrared video data, cutter temperature data, and rock debris image data may include:

[0046] A pre-built multi-source data acquisition system is used to acquire and store real-time tunneling parameter data, cutter vibration data, cutter infrared video data, cutter temperature data, and rock debris image data. The multi-source data acquisition system includes data acquisition hardware devices, a data acquisition gateway, and a server. The data acquisition hardware devices transmit data to the data acquisition gateway using edge computing interface programs or digital drivers. The data acquisition gateway uses various acquisition programs to remotely transmit the data to the server. The server receives the data through a message queue and stores the data in a Redis database or a MongoDB database.

[0047] Figure 2 This is a schematic diagram of a multi-source data acquisition system in an embodiment of the present invention, such as... Figure 2As shown, a multi-source data acquisition system architecture was constructed for multi-modal data from PLC (Programmable Logic Controller) devices, a rock cuttings detection system, and cutterhead infrared temperature flow. The data acquisition hardware includes a PLC device, a cutterhead infrared monitoring device, a rock cuttings image device, a geological advanced detection device, and a vibration detection device. The PLC device transmits data to the data acquisition gateway via a data driver program, and then to the data acquisition server via the PLC data source acquisition program and the remote data transmission program. The cutterhead infrared monitoring device transmits data to the data acquisition gateway via a data driver program, and then to the data acquisition server via the cutterhead infrared data source acquisition program and the remote data transmission program. The rock cuttings image device transmits data to the data acquisition gateway via an edge computing interface program, and then to the data acquisition server via the rock cuttings data source acquisition program and the remote data transmission program. The geological advanced detection device transmits data to the data acquisition gateway via an edge computing interface program, and then to the data acquisition server via the geological advanced detection data source acquisition program and the remote data transmission program. The vibration detection device transmits data to the data acquisition gateway via a data interface program, and then to the data acquisition server via the vibration data source acquisition program and the remote data transmission program. During implementation, data can also be collected through other devices to build a self-model and perform fusion processing. The data acquisition server's remote data receiving program receives the transmitted data, stores it in a message queue for processing, and finally stores the data in a Redis or MongoDB database using either a Redis or MongoDB storage program. This multi-source data acquisition system establishes a multi-type data storage mapping table, forming a big data storage warehouse with MongoDB as its core and utilizing multiple database technologies. The hardware system is based on an ARM (Advanced RISC Machine) microprocessor to achieve intelligent data acquisition for TBMs (Tunnel Boring Machines).

[0048] In one embodiment, real-time tunneling parameter data, cutter vibration data, cutter infrared video data, cutter temperature data, and rock debris image data are preprocessed. The preprocessing includes moving average filtering and outlier removal to improve data quality.

[0049] In step 102, historical tunneling parameter data is obtained according to a specified mileage. The historical tunneling parameter data and real-time tunneling parameter data are input into the first cutter head damage early warning model, and the first early warning result is output. The first cutter head damage early warning model is used to: perform trend comparison analysis on the historical tunneling parameter data, real-time tunneling parameter data and historical tunneling parameters when the cutter head is abnormal, and determine the first early warning result based on the trend comparison analysis result.

[0050] For example, real-time tunneling parameter data includes current location information. In order to analyze trends, historical tunneling parameter data prior to the real-time tunneling parameter data is obtained at a specified mileage. The historical tunneling parameter data and the real-time tunneling parameter data are combined to analyze the changing trends of some features. In this case, the changing trends of the same features of the tunneling parameters when the cutter was abnormal or normal in the past are analyzed in advance. The two (two changing trends) are compared to determine the current cutter status reflected by the real-time tunneling parameter data.

[0051] In one embodiment, the first hob damage early warning model is specifically used for:

[0052] The tunneling features of the defective data are extracted from the tunneling parameters during historical cutter anomalies; the tunneling features include root mean square value, peak value of the spectrum, and first waveform parameters;

[0053] Extract current tunneling features from data composed of historical tunneling parameter data and real-time tunneling parameter data using a set sliding window;

[0054] A trend comparison analysis is performed between the current tunneling characteristics and the tunneling characteristics of adverse data, and the first warning result is determined based on the trend comparison analysis results.

[0055] For example, for historical stable data sections, tunneling parameter data is filtered by mileage, and tunneling features such as root mean square value, peak frequency, and waveform indicators are extracted to obtain the data trends of healthy and unhealthy data and the degree of deviation from the normal distribution. For real-time tunneling parameter data collected by equipment, historical data is extracted backward by a certain mileage, discarding data with large deviations. Feature extraction and filtering are performed on other data within the sliding window, and feature data similar to historical data is processed. For data with similar trends to unhealthy data, cutter head damage is determined.

[0056] Figure 3 This is a schematic diagram of tunneling parameter data feature analysis in an embodiment of the present invention, such as... Figure 3 As shown, this graph represents the trend of the cutter head torque over time, where the horizontal axis represents time, the left vertical axis represents the variance of the cutter head torque, and the right vertical axis represents the cutter head torque.

[0057] In step 103, the hob vibration data is input into the second hob damage early warning model, and the second early warning result is output; the second hob damage early warning model is trained in advance using historical hob vibration data and hob damage status data to train a gated recurrent neural network.

[0058] During implementation, historical data on hob vibration and hob damage status are collected to form training and testing sets. A gated recurrent neural network is constructed, trained using the training set, and tested using the testing set, ultimately yielding the second hob damage early warning model.

[0059] In one embodiment, the second hob damage early warning model can also be obtained as follows:

[0060] Vibration signal features are extracted from historical hobbing vibration data; the vibration signal features include one or any combination of second waveform parameters, variance, peak-to-peak value, peak index, spectral bandwidth, and spectral centroid.

[0061] Based on the vibration signal characteristics and historical hob damage status data, the mapping relationship between vibration signal characteristics and hob damage status is determined; the hob damage status includes one or any combination of abnormal hob wear, hob chipping, and hob wear to the limit.

[0062] By utilizing the mapping relationship between vibration signal characteristics and cutter damage state, a gated recurrent neural network is trained to obtain a second cutter damage early warning model.

[0063] For example, vibration signal features such as waveform indices, variance, peak-to-peak value, peak index, spectral bandwidth, and spectral centroid are extracted from vibration data. The one-to-one correspondence between abnormal vibration data and hob damage states such as abnormal hob wear, hob chipping, and hob wear to completion is analyzed. For instance, when a hob is abnormally worn, the spectral centroid of the vibration signal may shift, and the variance will also change significantly; hob chipping will generate sharp pulses in the time-domain waveform, with an increased peak-to-peak value. Based on the above relationships, the collected historical vibration data is calibrated and input into a gated recurrent neural network (LSTM) model for training. The LSTM network, based on the recurrent neural network (RNN), introduces gating devices (input gate, forget gate, and output gate) to selectively memorize and update information, effectively avoiding the gradient vanishing and gradient exploding problems common in RNNs, and can better handle long-sequence vibration data. In terms of specific architecture, each LSTM unit contains four key components: cell state, input gate, forget gate, and output gate. An attention mechanism is introduced to make the model pay more attention to the key vibration features for identifying the damage state of the hob during the training process, thereby improving the model's early warning accuracy and generalization ability. Finally, a second hob damage early warning model based on vibration data is constructed.

[0064] Figure 4 This is a schematic diagram of vibration data feature analysis in an embodiment of the present invention, such as... Figure 4 As shown, the peak vibration value of the cutterhead varies with time for different projects, where the horizontal axis represents the vibration sampling time and the vertical axis represents the peak vibration value at the cutterhead.

[0065] In step 104, the cutter infrared video data and cutter temperature data are input into the third cutter damage early warning model, and the third early warning result is output. The third cutter damage early warning model is used to determine the third early warning result by using the pre-established mapping relationship between the cutter damage state and temperature value during cutter operation, the cutter infrared video data, and the cutter temperature data.

[0066] Infrared cameras can acquire video stream data and temperature data. By analyzing the relationship between temperature and cutter damage status, a third early warning result can be determined.

[0067] Figure 5 This is a schematic diagram illustrating the relationship between the damage state of the hob and temperature data in an embodiment of the present invention, as shown below. Figure 5 As shown, the temperature of each cutter within a certain tunneling distance is represented by the cutter number on the horizontal axis and the temperature on the vertical axis.

[0068] During implementation, an infrared sensor is installed behind the cutter head. The infrared sensor captures images of the cutter head rotating, and the images obtained by frame extraction from the captured video stream are as follows: Figure 5 As shown, the hobbing cutter generates a large amount of heat when cutting rock strata, resulting in the hobbing cutter area being in a state of extreme heat compared to other parts of the cutterhead. Analysis of hobbing damage and temperature changes reveals that when the hobbing cutter experiences uneven wear or chipping, the damaged hobbing cutter exhibits abnormally high or low temperatures compared to surrounding hobbing cutters. Based on these patterns, a third hobbing cutter damage early warning model is constructed, taking the acquired video stream and temperature stream of the cutterhead rotation as input and outputting the third early warning result.

[0069] In step 105, the rock slag image data is input into the fourth cutter damage early warning model, and the fourth early warning result is output. The fourth cutter damage early warning model is used to: determine the current rock slag feature value based on the rock slag image data, and determine the fourth early warning result using the pre-established mapping relationship between the cutter damage state and rock slag features during cutter operation and the current rock slag feature value.

[0070] During tunneling, the cutter head breaks the surrounding rock into rock debris. By identifying the rock debris, we can obtain information about the rock's hardness, weathering characteristics, and other geometric features. Establishing a mapping relationship between the cutter head's damage state and the characteristics of the rock debris allows us to determine the cutter head's damage state using these characteristics.

[0071] In one embodiment, the fourth hob damage early warning model is specifically used for:

[0072] Based on a pre-constructed slag fragment image segmentation model, the current slag feature values ​​are determined from the slag image data; the current slag feature values ​​include particle size distribution data and large slag fragment variation trend data.

[0073] The fourth early warning result is determined by using the mapping relationship between the cutter damage state and rock debris characteristics during cutter operation, which is pre-established based on Spearman rank correlation, and the current rock debris characteristic value.

[0074] For example, by installing a laser camera above the conveyor belt to acquire images of TBM slag fragments, and using deep learning-based slag fragment image segmentation technology, the particle size distribution of the TBM slag fragment images is obtained. The influence of rock slag particle distribution and mineral composition on cutter wear indices is analyzed using the Spearman rank correlation index, ultimately resulting in a fourth cutter damage early warning model with the proportion of large rock slag fragments, uniformity coefficient, and rock wear index as influencing factors.

[0075] Figure 6 This is a schematic diagram of slag image analysis in an embodiment of the present invention. Figure 1 The image shows the slag after the hob chipped. Figure 7 This is a schematic diagram of slag image analysis in an embodiment of the present invention. Figure 2 The graph shows the cumulative distribution curve of slag fragments, where the horizontal axis represents particle size and the vertical axis represents the proportion of large slag pieces.

[0076] Step 106: Based on Bayes' theorem, the first warning result, the second warning result, the third warning result and the fourth warning result are integrated to obtain the hob state warning result. The hob state warning result includes hob states such as hob uneven wear, hob chipping, and wear reaching the limit.

[0077] For example, by using Bayes' theorem and setting posterior probabilities, the warning results of each sub-model can be organically integrated, fully considering the correlation between each sub-model, to achieve multi-information fusion for judging abnormal wear of hobs, significantly improving the accuracy of hob abnormal wear warning. This method can be applied to various complex working conditions and multiple devices, and has important guiding significance for the field.

[0078] In summary, this embodiment of the invention takes into account that damage to the cutter head hob can lead to a series of adverse reactions such as abnormal high temperature and undesirable vibration. It constructs early warning models for each modal data collected, assigns different weights to each model during fusion processing, and performs final decision-level fusion of the strategies output by each model to finally obtain the hob status early warning result. The model is constructed in four parallel sub-models: First, it processes tunneling data such as propulsion speed, propulsion force, cutterhead rotation speed, and cutterhead torque to obtain the values ​​of corresponding characteristic data, summarizes the changing patterns of characteristic data when the cutterhead is damaged, and constructs a first cutterhead damage early warning model based on tunneling parameters. Second, it acquires vibration data from vibration sensors installed on the cutterhead, extracts a series of vibration data features such as kurtosis, peak frequency, pulse index, and waveform index, compares and analyzes the distribution patterns of healthy and abnormal features, and constructs a second cutterhead damage early warning model based on vibration data. Third, it collects video of cutterhead rotation from an infrared camera installed behind the cutterhead and obtains the corresponding temperature flow, analyzes the abnormal temperature change patterns of the cutterhead, and constructs a third cutterhead damage early warning model. Fourth, it detects rock debris through the conveyor belt slag system, thereby inferring the specific conditions of the surrounding rock and obtaining the rock mass strength. When tunneling in hard rock strata, the abnormal high temperature generated by cutterhead wear or chipping is more obvious. Based on the above patterns, it constructs a relationship model between rock mass strength and cutterhead damage—a fourth cutterhead damage early warning model. Finally, using Bayes' theorem, combined with prior probability and likelihood function, the decision results of the above independent models are fused to achieve a comprehensive assessment of hob damage based on multi-source information.

[0079] Compared with existing technologies, this invention provides a method for early warning of abnormal cutter conditions based on multimodal information fusion. It comprehensively considers various factors such as vibration data, tunneling parameters, rock mass strength, and cutter temperature, and constructs an early warning model for abnormal cutter conditions based on multimodal fusion technology. This avoids the problem of early warning failure caused by the damage of a single early warning device, improves the accuracy of early warning, provides a basis for on-site cutter replacement personnel, and reduces cutter replacement costs.

[0080] This invention also provides a tunneling machine cutter abnormality warning device, as described in the following embodiments. Since the principle by which this device solves the problem is similar to the tunneling machine cutter abnormality warning method, the implementation of this device can refer to the implementation of the tunneling machine cutter abnormality warning method; repeated details will not be elaborated further.

[0081] Figure 8 This is a schematic diagram of the early warning device for abnormal status of the tunneling machine cutter head in an embodiment of the present invention, as shown below. Figure 7 As shown, the device includes:

[0082] The data acquisition module 801 is used to acquire real-time tunneling parameter data, cutter vibration data, cutter infrared video data, cutter temperature data, and rock debris image data.

[0083] The first sub-model processing module 802 is used to acquire historical tunneling parameter data according to a specified mileage, input the historical tunneling parameter data and real-time tunneling parameter data into the first cutter head damage early warning model, and output the first early warning result; the first cutter head damage early warning model is used to: perform trend comparison analysis on the historical tunneling parameter data, real-time tunneling parameter data and historical tunneling parameters when the cutter head is abnormal, and determine the first early warning result based on the trend comparison analysis result;

[0084] The second sub-model processing module 803 is used to input the hob vibration data into the second hob damage early warning model and output the second early warning result; the second hob damage early warning model is pre-trained on a gated recurrent neural network using historical hob vibration data and hob damage status data.

[0085] The third sub-model processing module 804 is used to input the cutter infrared video data and cutter temperature data into the third cutter damage early warning model and output the third early warning result; the third cutter damage early warning model is used to: determine the third early warning result by using the pre-established mapping relationship between the cutter damage state and temperature value during cutter operation, the cutter infrared video data and the cutter temperature data.

[0086] The fourth sub-model processing module 805 is used to input rock slag image data into the fourth cutter damage early warning model and output the fourth early warning result; the fourth cutter damage early warning model is used to: determine the current rock slag feature value based on the rock slag image data, and determine the fourth early warning result using the pre-established mapping relationship between the cutter damage state and rock slag features during cutter operation and the current rock slag feature value;

[0087] The fusion processing module 806 is used to fuse the first warning result, the second warning result, the third warning result and the fourth warning result based on Bayes' theorem to obtain the rolling cutter status warning result.

[0088] In one embodiment, the data acquisition module 801 is specifically used for:

[0089] A pre-built multi-source data acquisition system is used to acquire and store real-time tunneling parameter data, cutter vibration data, cutter infrared video data, cutter temperature data, and rock debris image data. The multi-source data acquisition system includes data acquisition hardware devices, a data acquisition gateway, and a server. The data acquisition hardware devices transmit data to the data acquisition gateway using edge computing interface programs or digital drivers. The data acquisition gateway uses various acquisition programs to remotely transmit the data to the server. The server receives the data through a message queue and stores the data in a Redis database or a MongoDB database.

[0090] In one embodiment, the first hob damage early warning model is specifically used for:

[0091] The tunneling features of the defective data are extracted from the tunneling parameters during historical cutter anomalies; the tunneling features include root mean square value, peak value of the spectrum, and first waveform parameters;

[0092] Extract current tunneling features from data composed of historical tunneling parameter data and real-time tunneling parameter data using a set sliding window;

[0093] A trend comparison analysis is performed between the current tunneling characteristics and the tunneling characteristics of adverse data, and the first warning result is determined based on the trend comparison analysis results.

[0094] In one embodiment, the second hob damage early warning model is obtained as follows:

[0095] Vibration signal features are extracted from historical hobbing vibration data; the vibration signal features include one or any combination of second waveform parameters, variance, peak-to-peak value, peak index, spectral bandwidth, and spectral centroid.

[0096] Based on the vibration signal characteristics and historical hob damage status data, the mapping relationship between vibration signal characteristics and hob damage status is determined; the hob damage status includes one or any combination of abnormal hob wear, hob chipping, and hob wear to the limit.

[0097] By utilizing the mapping relationship between vibration signal characteristics and cutter damage state, a gated recurrent neural network is trained to obtain a second cutter damage early warning model.

[0098] In one embodiment, the fourth hob damage early warning model is specifically used for:

[0099] Based on a pre-constructed slag fragment image segmentation model, the current slag feature values ​​are determined from the slag image data; the current slag feature values ​​include particle size distribution data and large slag fragment variation trend data.

[0100] The fourth early warning result is determined by using the mapping relationship between the cutter damage state and rock debris characteristics during cutter operation, which is pre-established based on Spearman rank correlation, and the current rock debris characteristic value.

[0101] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described method for early warning of abnormal conditions of the tunneling machine cutter head.

[0102] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for early warning of abnormal conditions of the tunneling machine cutter head.

[0103] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described method for early warning of abnormal conditions of the tunneling machine cutter head.

[0104] In this embodiment of the invention, real-time tunneling parameter data, cutter vibration data, cutter infrared video data, and rock debris image data are acquired, and four sub-models are constructed. Each sub-model is used to process and analyze the real-time tunneling parameter data, cutter vibration data, cutter infrared video data, and rock debris image data. Finally, the warning results of the four sub-models are combined for fusion judgment. This achieves comprehensive consideration of multi-scale and multi-directional cutter damage factors, avoiding the problem of warning failure caused by the failure of a single warning device. Single data corruption or sudden changes will not have a significant impact on the warning results, greatly improving the accuracy of cutter abnormality warnings and realizing comprehensive monitoring and warning of cutter status.

[0105] 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 embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0106] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0107] 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.

[0108] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0109] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for early warning of abnormal conditions of tunneling machine cutterheads, characterized in that, include: Acquire real-time tunneling parameter data, cutter vibration data, cutter infrared video data, cutter temperature data, and rock debris image data; Historical tunneling parameter data is acquired according to a specified mileage. The historical tunneling parameter data and real-time tunneling parameter data are input into the first cutter head damage early warning model, and the first early warning result is output. The first cutter head damage early warning model is used to: perform trend comparison analysis on the historical tunneling parameter data, real-time tunneling parameter data and historical tunneling parameters when the cutter head is abnormal, and determine the first early warning result based on the trend comparison analysis result. Input the hob vibration data into the second hob damage early warning model and output the second early warning result; The second hob damage early warning model is obtained by training a gated recurrent neural network using historical hob vibration data and hob damage status data. The cutter infrared video data and cutter temperature data are input into the third cutter damage early warning model, and the third early warning result is output. The third cutter damage early warning model is used to: determine the third early warning result by using the pre-established mapping relationship between the cutter damage state and temperature value during cutter operation, the cutter infrared video data and the cutter temperature data. The rock slag image data is input into the fourth cutter damage early warning model, and the fourth early warning result is output. The fourth cutter damage early warning model is used to: determine the current rock slag feature value based on the rock slag image data, and determine the fourth early warning result using the pre-established mapping relationship between the cutter damage state and rock slag features during cutter operation and the current rock slag feature value. Based on Bayes' theorem, the first, second, third, and fourth warning results are merged to obtain the rolling cutter status warning result.

2. The method as described in claim 1, characterized in that, Acquire real-time tunneling parameter data, cutter vibration data, cutter infrared video data, cutter temperature data, and rock debris image data, including: A pre-built multi-source data acquisition system is used to acquire and store real-time tunneling parameter data, cutter vibration data, cutter infrared video data, cutter temperature data, and rock debris image data. The multi-source data acquisition system includes data acquisition hardware devices, a data acquisition gateway, and a server. The data acquisition hardware devices transmit data to the data acquisition gateway using edge computing interface programs or digital drivers. The data acquisition gateway uses various acquisition programs to remotely transmit the data to the server. The server receives the data through a message queue and stores the data in a Redis database or a MongoDB database.

3. The method as described in claim 1, characterized in that, The first hob damage early warning model is specifically used for: The tunneling features of the defective data are extracted from the tunneling parameters during historical cutter anomalies; the tunneling features include root mean square value, peak value of the spectrum, and first waveform parameters; Extract current tunneling features from data composed of historical tunneling parameter data and real-time tunneling parameter data using a set sliding window; A trend comparison analysis is performed between the current tunneling characteristics and the tunneling characteristics of adverse data, and the first warning result is determined based on the trend comparison analysis results.

4. The method as described in claim 1, characterized in that, The second hob damage early warning model is obtained as follows: Vibration signal features are extracted from historical hobbing vibration data; the vibration signal features include one or any combination of second waveform parameters, variance, peak-to-peak value, peak index, spectral bandwidth, and spectral centroid. Based on the vibration signal characteristics and historical hob damage status data, the mapping relationship between vibration signal characteristics and hob damage status is determined; the hob damage status includes one or any combination of abnormal hob wear, hob chipping, and hob wear to the limit. By utilizing the mapping relationship between vibration signal characteristics and cutter damage state, a gated recurrent neural network is trained to obtain a second cutter damage early warning model.

5. The method as described in claim 1, characterized in that, The fourth hob damage early warning model is specifically used for: Based on a pre-constructed slag fragment image segmentation model, the current slag feature values ​​are determined from the slag image data; the current slag feature values ​​include particle size distribution data and large slag fragment variation trend data. The fourth early warning result is determined by using the mapping relationship between the cutter damage state and rock debris characteristics during cutter operation, which is pre-established based on Spearman rank correlation, and the current rock debris characteristic value.

6. A tunneling machine cutter abnormality early warning device, characterized in that, include: The data acquisition module is used to acquire real-time tunneling parameter data, cutter vibration data, cutter infrared video data, cutter temperature data, and rock debris image data; The first sub-model processing module is used to acquire historical tunneling parameter data according to a specified mileage, input the historical tunneling parameter data and real-time tunneling parameter data into the first cutter head damage early warning model, and output the first early warning result; the first cutter head damage early warning model is used to: perform trend comparison analysis on the historical tunneling parameter data, real-time tunneling parameter data and historical tunneling parameters when the cutter head is abnormal, and determine the first early warning result based on the trend comparison analysis result; The second sub-model processing module is used to input the hob vibration data into the second hob damage early warning model and output the second early warning result. The second hob damage early warning model is obtained by training a gated recurrent neural network using historical hob vibration data and hob damage status data. The third sub-model processing module is used to input the cutter infrared video data and cutter temperature data into the third cutter damage early warning model and output the third early warning result. The third cutter damage early warning model is used to: determine the third early warning result by using the pre-established mapping relationship between the cutter damage state and temperature value during cutter operation, the cutter infrared video data and the cutter temperature data. The fourth sub-model processing module is used to input rock slag image data into the fourth cutter damage early warning model and output the fourth early warning result; the fourth cutter damage early warning model is used to: determine the current rock slag feature value based on the rock slag image data, and determine the fourth early warning result using the pre-established mapping relationship between the cutter damage state and rock slag features during cutter operation and the current rock slag feature value; The fusion processing module is used to fuse the first, second, third, and fourth warning results based on Bayes' theorem to obtain the rolling cutter status warning result.

7. The apparatus as claimed in claim 6, characterized in that, The data acquisition module is specifically used for: A pre-built multi-source data acquisition system is used to acquire and store real-time tunneling parameter data, cutter vibration data, cutter infrared video data, cutter temperature data, and rock debris image data. The multi-source data acquisition system includes data acquisition hardware devices, a data acquisition gateway, and a server. The data acquisition hardware devices transmit data to the data acquisition gateway using edge computing interface programs or digital drivers. The data acquisition gateway uses various acquisition programs to remotely transmit the data to the server. The server receives the data through a message queue and stores the data in a Redis database or a MongoDB database.

8. The apparatus as claimed in claim 6, characterized in that, The first hob damage early warning model is specifically used for: The tunneling features of the defective data are extracted from the tunneling parameters during historical cutter anomalies; the tunneling features include root mean square value, peak value of the spectrum, and first waveform parameters; Extract current tunneling features from data composed of historical tunneling parameter data and real-time tunneling parameter data using a set sliding window; A trend comparison analysis is performed between the current tunneling characteristics and the tunneling characteristics of adverse data, and the first warning result is determined based on the trend comparison analysis results.

9. The apparatus as claimed in claim 6, characterized in that, The second hob damage early warning model is obtained as follows: Vibration signal features are extracted from historical hobbing vibration data; the vibration signal features include one or any combination of second waveform parameters, variance, peak-to-peak value, peak index, spectral bandwidth, and spectral centroid. Based on the vibration signal characteristics and historical hob damage status data, the mapping relationship between vibration signal characteristics and hob damage status is determined; the hob damage status includes one or any combination of abnormal hob wear, hob chipping, and hob wear to the limit. By utilizing the mapping relationship between vibration signal characteristics and cutter damage state, a gated recurrent neural network is trained to obtain a second cutter damage early warning model.

10. The apparatus as claimed in claim 6, characterized in that, The fourth hob damage early warning model is specifically used for: Based on a pre-constructed slag fragment image segmentation model, the current slag feature values ​​are determined from the slag image data; the current slag feature values ​​include particle size distribution data and large slag fragment variation trend data. The fourth early warning result is determined by using the mapping relationship between the cutter damage state and rock debris characteristics during cutter operation, which is pre-established based on Spearman rank correlation, and the current rock debris characteristic value.

11. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 5.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 5.

13. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 5.

Images