Tunnel surrounding rock stability diagnosis method based on tmb multi-modal data fusion

Abstract

The application discloses a tunnel surrounding rock stability diagnosis method based on TBM multi-modal data fusion, which comprises the following steps: 1, dividing the surrounding rock stability type; 2, TBM multi-modal data preparation; 3, TBM multi-modal data space-time alignment and fusion; 4, multi-modal data null and repeated value filling; 5, multi-modal data feature extraction based on the excavation section; 6, constructing a TBM multi-modal feature database based on the excavation section; 7, constructing a tunnel surrounding rock stability prediction model; 8, tunnel surrounding rock stability field diagnosis application. The application can realize the repair and completion of the low and high null degree modal data through the nearest neighbor null filling and linear interpolation method based on the time and space information to ensure the accuracy of the alignment between different modal data, can be applied to the fusion of different monitoring modes in the tunnel engineering construction process, can realize the rapid diagnosis of the field surrounding rock stability, and can ensure the safety of the construction environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of tunnel boring machine construction monitoring information, in particular to a tunnel surrounding rock stability diagnosis method based on TBM multi-modal data fusion. BACKGROUND

[0002] The tunnel boring machine excavation method has the advantages of high mechanization degree, fast excavation speed and safe construction environment, so it is widely used in tunnel engineering excavation. In the construction process, due to the constraints of the working space, traditional advanced geological exploration methods such as microseismic, geological radar and other methods have the problems of high cost, affecting the normal excavation process and the diagnosis results are not timely. This leads to the diagnosis of the stability of the surrounding rock during the mechanized construction of the tunnel mainly relying on the preliminary geological exploration report, however, the tunnel boring machine is sensitive to changes in geological conditions, and when encountering extremely broken and unstable sections, it may cause major engineering accident disasters such as collapse and machine jamming, which seriously affects the safety of life and property of the on-site workers. Therefore, there is an urgent need to develop a real-time prediction method for on-site surrounding rock stability diagnosis.

[0003] In recent years, intelligent prediction models for surrounding rock stability during construction based on on-site monitoring data have been developed. Most existing methods are based on the same modal type data source for data processing, feature extraction and model construction, which has the disadvantages of single data source, high correlation and lack of complementarity, which further leads to limited model prediction performance. However, with the advancement of sensing technology, it has become possible to monitor and acquire different modal data during the TBM excavation process. For example, tunnel boring machine operating parameters, cutter bit radial wear values at different cutter positions on the cutter head, vibration data at different monitoring points on the cutter head, and images of slag pieces on the surrounding rock or belt conveyor during construction are all collected and stored during the tunnel excavation process. Based on the fusion processing of different modal data during the machine excavation process, a real-time diagnosis model for surrounding rock stability can be established, which can greatly reduce the cost of advanced geological exploration on site, accurately predict the stability of the surrounding rock, and realize rapid diagnosis and early warning of unstable sections, thereby ensuring safe and efficient excavation process on site.

[0004] To realize real-time diagnosis of surrounding rock stability based on multi-modal data, detailed construction data of different modal dimensions under different geological conditions need to be collected, and a correlation fusion database of different data sources needs to be constructed. A standardized fusion processing technology system for multi-source data needs to be established to realize multi-modal data alignment and completion, and reasonable frequency feature parameters need to be extracted for the tunnel boring process. This provides an important data basis for the construction of a real-time diagnosis model for surrounding rock stability, which can not only realize rapid diagnosis of surrounding rock, but also reduce the time processing cost of high-frequency modal data. This method can effectively reduce the cost of advanced geological exploration on site, efficiently and accurately evaluate the state of the surrounding rock during the construction process of the tunnel boring machine, and thus ensure the safety of the construction environment. SUMMARY

[0005] The technical problem solved by the present application is to provide a tunnel surrounding rock stability diagnosis method based on TBM multi-modal data fusion to solve the problems of the prior art.

[0006] To solve the above technical problems, the technical scheme adopted by the present application is:

[0007] A tunnel surrounding rock stability diagnosis method based on TBM multi-modal data fusion comprises the following steps.

[0008] Step 1, dividing the surrounding rock stability type: the tunnel surrounding rock is divided into five types of stable, basically stable, locally stable, unstable and extremely unstable.

[0009] Step 2, TBM multi-modal data preparation: for each type of stability state type tunnel surrounding rock, TBM mechanized construction method is used for tunneling, and different surrounding rock stability type modal data is continuously monitored and collected during tunneling construction; thus, n kinds of modal data are collected under each type of surrounding rock stability state type working condition; then, different modal data under each type of stability state type working condition is divided into n kinds of modal data blocks in units of days.

[0010] Step 3, TBM multi-modal data space-time alignment: for each type of stability state type tunnel surrounding rock n kinds of modal data blocks, the sampling frequency is sorted from high to low, and the modal data block corresponding to the highest sampling frequency is taken as the reference, and the remaining n-1 kinds of modal data blocks are copied into the modal data block corresponding to the highest sampling frequency, to realize alignment and fusion processing.

[0011] Step 4, data filling: identify the null and duplicate values in the TBM multi-modal data after time and space alignment in step 4, and repair the identified null and duplicate values by using the nearest neighbor filling or linear interpolation method.

[0012] Step 5, feature extraction: for each type of stability state type tunnel surrounding rock data filling n kinds of modal data blocks, the feature extraction is carried out in units of tunneling sections; a plurality of values are extracted as feature values for each modal data in a tunneling section, and n in features, then n sum =n1+n2+…+n n n features; wherein the tunneling section refers to the stage data from zero to non-zero and finally to zero of the cutterhead rotation speed of the TBM from the cutterhead start to the maximum stroke shutdown.

[0013] Step 6, constructing a TBM multi-modal feature database based on the tunneling section: based on all the feature values of all the tunneling sections extracted in step 5 under different surrounding rock stability working conditions, a TBM multi-modal feature database based on the tunneling section is constructed.

[0014] Step 7, constructing a tunnel surrounding rock stability prediction model: the tunnel surrounding rock stability prediction model is a machine learning model, the input of the tunnel surrounding rock stability prediction model is the feature values of n modal data of a tunneling section, and the output of the tunnel surrounding rock stability prediction model is the surrounding rock stability type corresponding to the tunneling section. The tunnel surrounding rock stability prediction model is trained using the TBM multi-modal feature database based on the tunneling section constructed in step 6.

[0015] Step 8, tunnel surrounding rock stability diagnosis application: monitoring n sum modal data of the TBM in the tunneling construction process, and obtaining n sum feature values of the current tunneling section under different modalities; then, inputting the n sum feature values of the current tunneling section into the tunnel surrounding rock stability prediction model trained in step 7 to predict the surrounding rock stability state of the current tunneling section in the construction process.

[0016] In step 2, the modal data includes TBM operation data, cutterhead cutter wear data, cutterhead vibration data, and surrounding rock or spoil image.

[0017] In step 2, the TBM operation data includes thrust, torque, advancing speed, rotation speed and penetration; the cutterhead cutter wear data includes the radial wear data of each cutter position cutter on the cutterhead; and the cutterhead vibration data includes the vibration acceleration data at the monitoring position.

[0018] In step 3, the time and space alignment method of TBM multi-modal data includes the following steps:

[0019] Step 3-1, inter-modal sorting: sorting the n modal data blocks after data segmentation from high to low according to the sampling frequency, respectively: first modal data block, second modal data block, …, n modal data block.

[0020] Step 3-2, intra-modal sorting: sorting each modal data block according to the monitoring time or space stake position.

[0021] Step 3-3, second modality data block alignment: according to the time and space stake position, taking the time information of the first modality data block as the reference, through time indexing, inserting the data of each acquisition frequency in the second modality data block into the first modality data block with the same time precision, realizing the data alignment and fusion of the second modality data block and the first modality data block.

[0022] Step 3-4, repeating step 3-3 until the data alignment and fusion of the nth modality data block and the first modality data block are completed.

[0023] In step 4, the degree type of the null value and the repeated value needs to be determined first; for the modality data block with small degree of null value and repeated value, KNN nearest neighbor filling method is used to repair the null value and repeated value; for the modality data block with large degree of null value and repeated value, nearest neighbor linear interpolation method is used to repair the null value and repeated value.

[0024] In step 4, the specific method of data filling includes the following steps:

[0025] Step 4-1, identifying null value and repeated value: identifying the null value and repeated value in the second modality data block after alignment and fusion processing.

[0026] Step 4-2, determining the degree of null value and repeated value: comparing the number of null value and repeated value of each modality in each day with the set proportion value, when the number of null value and repeated value in the ith day is less than the set proportion value, all null value and repeated value in the ith day are judged as small degree of null value or repeated value, otherwise, it is judged as large degree of null value or repeated value.

[0027] Step 4-3, null value and repeated value repair: when the degree of null value or repeated value in the second modality data block determined in step 4-2 is small, KNN nearest neighbor filling method is used for repair; when the degree of null value or repeated value in the second modality data block determined in step 4-2 is large, nearest neighbor linear interpolation method is used for repair, as shown in Figure 3

[0028] Step 4-4, repeating step 4-1 to step 4-3, sequentially completing the null value and repeated value repair of the third modality data block to the nth modality data block.

[0029] In step 5, according to the form of tunneling section, different modality data are respectively extracted by statistical characteristics (such as mean value) or physical index extraction method, as shown in Figure 4 For example, the ith modality in the same tunneling section obtains n i characteristics, then for all n kinds of modality data blocks, n sum = n1+n2+…+n n ​characteristic values; wherein the tunneling section refers to the stage data in which the cutterhead rotating speed of the TBM changes from zero to a non-zero value and finally decreases to zero from the cutterhead start to the maximum stroke shutdown process.

[0030] In step 7, the deep learning model is a prediction model based on gradient boosting ensemble learning algorithm.

[0031] In step 4-2, the proportion threshold for judging null and duplicate value types is set to 5% or 10% of the total number of samples in the corresponding mode on the corresponding day.

[0032] The present application has the following beneficial effects:

[0033] (1) For the prediction of surrounding rock stability in the TBM tunnel construction process, the traditional method based on geological exploration and advanced geological drilling has high construction cost, long prediction and analysis time, and is not conducive to efficient and timely diagnosis in mechanized rapid construction. The present application adopts a multi-modal monitoring data construction method based on the tunnel construction site to realize the alignment of different modal data and reduce the time processing cost of high-frequency data, providing a basis for timely diagnosis of the stability of the surrounding rock on site, and further ensuring the safe and efficient construction process.

[0034] (2) The present application proposes a multi-modal data alignment method based on space-time ordering, which ensures the accuracy of data block alignment between different modalities from two dimensions of time information and spatial position, and realizes the effective fusion of multi-modal data.

[0035] (3) The multi-modal surrounding rock stability diagnosis model proposed by the present application can be applied to the timely diagnosis of the stability of the surrounding rock in the construction process based on the on-site monitoring data, significantly reducing the cost of on-site advanced geological drilling and other engineering site tests, while avoiding the impact on the tunnel boring machine construction, improving the tunnel excavation efficiency. The established diagnosis model can also provide tunnel instability risk warning, thereby greatly avoiding the occurrence of major engineering accidents and ensuring construction safety. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 The flowchart of the tunnel surrounding rock stability diagnosis method based on TBM multi-modal data fusion of the present application.

[0037] Figure 2 The fusion data structure after alignment based on space-time ordering for TBM monitoring modal 1, modal 2, modal n and their own data structure arranged from high to low sampling frequency.

[0038] Figure 3Fig. (a) is a schematic diagram of the nearest neighbor KNN filling principle, mainly suitable for repairing modal data with low degree of null value; Fig. (b) is a schematic diagram of the nearest neighbor linear interpolation filling principle, mainly suitable for modal data with high degree of null value.

[0039] Figure 4 Fig. (a) shows a schematic diagram of feature extraction for different modal data of each driving section; and Fig. (b) shows a schematic diagram of a database after feature fusion of different modal data of each driving section.

[0040] Figure 5 Fig. is an example diagram of TBM vibration, machine and cutter wear modal fusion database of Yinhua-Ji'eliang.

[0041] Figure 6 Fig. is an example diagram of TBM vibration, machine and cutter wear modal fusion database of Yinhua-Ji'eliang.

[0042] Figure 7 Fig. is a surrounding rock stability classification prediction result based on TBM multi-modal data fusion of Yinhua-Ji'eliang. DETAILED DESCRIPTION

[0043] The present application will be further described in detail below in combination with the drawings and specific preferred embodiments.

[0044] The present application takes the multi-modal monitoring data of the open TBM construction process of the No. 6 section tunnel of Yinhua-Ji'eliang in Inner Mongolia as an example to explain the present application in detail.

[0045] As shown in Fig. (a) is a schematic diagram of the nearest neighbor KNN filling principle, mainly suitable for repairing modal data with low degree of null value; Fig. (b) is a schematic diagram of the nearest neighbor linear interpolation filling principle, mainly suitable for modal data with high degree of null value. Figure 1 The tunnel surrounding rock stability diagnosis method based on TBM multi-modal data fusion comprises the following steps.

[0046] Step 1, dividing the stability type of surrounding rock: according to the national standard "Code for Geological Exploration of Water Conservancy and Hydropower Engineering GB50487-2008" for evaluating the stability of the surrounding rock of the present tunnel engineering site, the tunnel surrounding rock is divided into five types according to the known standard, i.e. stable, basically stable, locally stable, unstable and extremely unstable.

[0047] Step 2, TBM multi-modal data preparation

[0048] A, data collection

[0049] For each type of stability state of the tunnel surrounding rock, TBM is used for driving, and different types of modal data are continuously monitored and collected during the tunnel driving construction; thus, n types of modal data M = [M1, M2,..., Mn] are collected for each type of stability working condition of the tunnel surrounding rock.n ]。

[0050] The modal data includes TBM operation data, cutterhead roller wear data, and cutterhead vibration data. The TBM operation data includes thrust, torque, propulsion speed, rotation speed, and penetration, which are measured by displacement, speed, and force sensors and other types of sensors. The cutterhead roller wear data includes the radial wear data of 34 rollers on the cutterhead, and the wear values are measured by the on-site cutterhead maintenance construction team. The cutterhead vibration data is measured by vibration sensors installed at the rear of the cutterhead.

[0051] In this embodiment, the cutterhead vibration, machine operation, and roller wear modal data in the pile number range of 66058.81 to 55855.69 of the Yuhuo Jiliao No. 6 branch tunnel TBM field monitoring data are collected, and the rock mass stability in this pile number section range includes four states of basic stability, local stability, instability, and extreme instability.

[0052] In this embodiment, the machine operation modal data in the pile number range of 66058.81 to 55855.69 is mainly obtained by sensors to obtain total thrust, total torque, propulsion speed, rotation speed, and other parameters, and the sampling frequency is preferably 1 Hz / s.

[0053] In this embodiment, the cutterhead wear data in the pile number range of 66058.81 to 55855.69 includes the radial wear values of the 34 rollers on the cutterhead, and the sampling frequency is preferably 1 Hz / day.

[0054] In this embodiment, the cutterhead vibration data in the pile number range of 66058.81 to 55855.69 mainly includes 6 sensors installed at the rear of the cutterhead, and the sampling frequency is preferably 240 Hz / s.

[0055] B. Data segmentation

[0056] Each modal data of the tunnel surrounding rock in each stability state type is segmented by day to form n modal data blocks. Taking the data of the first day as an example

[0057] In this embodiment, the three-dimensional modal data in the pile number range of 66058.81 to 55855.69 is segmented into data blocks by day by writing a time segmentation function. Taking the data of December 19, 2020 as an example, the data representation form is

[0058] Step 3, spatio-temporal alignment of TBM multi-modal data: for each type of n-modal data block of tunnel surrounding rock in a stable state, sort them from high to low according to the sampling frequency, take the cutter head vibration modal data block corresponding to the highest sampling frequency as the reference, rely on time and space stake position, traverse the remaining n-1 modal data blocks, and copy and paste them into the highest sampling frequency corresponding modal data block to realize alignment and fusion processing.

[0059] The above-mentioned spatio-temporal alignment method of TBM multi-modal data preferably comprises the following steps.

[0060] Step 3-1, inter-modal sorting: sort the n-modal data blocks after data segmentation from high to low according to the sampling frequency, respectively: first modal data block (also called modal 1), second modal data block (also called modal 2), …, n-th modal data block (also called modal n). That is, the sampling frequency of modal 1 is the highest.

[0061] In this embodiment, first sort the three modal data according to the size of the sampling frequency to get: vibration modal sampling frequency > machine running modal sampling frequency > cutter wear modal sampling frequency.

[0062] Step 3-2, intra-modal sorting: sort each modal data block according to the order of monitoring time and space stake position. For example, first sort the monitoring data of the first day in modal 1 Arrange the monitoring data under the same monitoring modal in this data block according to the order of time record and stake displacement information.

[0063] Then sort the monitoring data of the first day in modal 2 Similarly, sort the data in modal 2 according to the time and stake record information.

[0064] Next, sort the monitoring data of the first day in modal 3 Similarly, sort the data in modal 3 according to the time and stake record information.

[0065] In turn, sort the monitoring data of the first day in modal n Sort them according to the time and stake record information.

[0066] In this embodiment, the cutter head vibration data with the highest sampling frequency is selected, and the data of December 19, 2020 is sorted according to the time position to get the data shown in Table 1. During the construction process of the No. 6 tunnel of the Yinhua Jiliao No. 2 section, a total of 6 sensors are installed on the TBM cutter head, and the sampling frequency of each sensor is 240 Hz / s, so the data of the first sensor from 1 to 240 Hz is shown as vibration 1-240 column data.

[0067] Table 1 Cutter head vibration modal data of Yinhua Jiliao No. 6 section

[0068]

[0069] Select the machine operation data of the second sampling frequency, and sort it according to the data of one day of time 2020 / 12 / 19 or the position of the stake number in order, and obtain the data shown in Table 2.

[0070] Table 2 Machine operation modal data of No. 6 bid section of Yujiajiliao

[0071]

[0072] Select the cutter wear data of the third sampling frequency, and sort it according to the time or the position of the stake number in order, and obtain the data shown in Table 3.

[0073] Table 3 Cutter radial wear value data (unit: mm) of different cutter positions of No. 6 bid section of Yujiajiliao

[0074]

[0075] Step 3-3, alignment of the second modal data block: according to the time and the stake number position, taking the time information of the first modal data block as the reference, the data of each collection frequency in the second modal data block is inserted into the first modal data block with the same time accuracy, so as to realize the data alignment and fusion of the second modal data block and the first modal data block.

[0076] Since the data monitoring frequency in modal 2 is less than the frequency of the data in modal 1, the method of realizing the fusion of the two modal data is to map the data in modal 2 to modal 1. Therefore, it is necessary to traverse the time and stake number index in modal 2, and after searching to the same space-time position in modal 1, the data in modal 2 is inserted into the corresponding position of modal 1 and a new column is added for storage. Finally, the space-time position alignment of the two modal data is realized.

[0077] That is, the sorted data in the modal is inserted into the corresponding position of modal 1 and a new column is added for storage. For example, according to the time information 2020 / 12 / 19 03:30:31 as the index, the data of the corresponding row of the machine operation modal data is extracted, and the corresponding data is copied and inserted into the corresponding cutter disc vibration modal data.

[0078] Step 3-4, repeat step 3-3 until the data alignment and fusion of the nth modal data block and the first modal data block at each time and stake number are completed; that is, the last modal data is respectively aligned in space-time with the modal 1 data with the highest sampling frequency, so as to realize the mapping relationship between the existing modal data and the highest sampling frequency modal 1.

[0079] Wherein, based on the time 2020 / 12 / 19, the stake position 63926.3 information, the wear data of different cutter positions on the same day is copied and inserted into the fusion data closest to the stake 63926.3 in the cutter disc vibration and machine fusion database, and a multi-modal fusion database is obtained as shown in Figure 5

[0080] Finally, steps 3-2 to 3-4 are repeated to complete the alignment and fusion of the multi-modal data of all working day data blocks.

[0081] Step 4, data filling: identifying the null values and duplicate values in the TBM multi-modal data after the time and space alignment in step 4, and repairing the identified null values and duplicate values by using the nearest neighbor filling or linear interpolation method.

[0082] When filling the data, the degree of the null values and duplicate values needs to be determined first; for the modal data blocks with small degree of null values and duplicate values, the KNN nearest neighbor filling method is used to repair the null values and duplicate values; for the modal data blocks with large degree of null values and duplicate values, the nearest neighbor linear interpolation method is used to repair the null values and duplicate values.

[0083] The specific method of the above data filling preferably includes the following steps.

[0084] Step 4-1, identifying null values and duplicate values: identifying the null values and duplicate values in the second modal data block after the alignment and fusion processing.

[0085] Step 4-2, determining the degree of null values and duplicate values: comparing the number of null values and duplicate values of each modal in each day with the set proportion value, when the number of null values and duplicate values in the ith day is less than the set proportion value, then all the null values and duplicate values in the ith day are judged as small degree of null values or duplicate values, otherwise, it is judged as large degree of null values or duplicate values.

[0086] The set proportion value is preferably 5% or 10% of the total number of the corresponding modal in the corresponding day.

[0087] Step 4-3, null value and duplicate value repair

[0088] When the degree of null values or duplicate values in the second modal data block determined in step 4-2 is small, the KNN nearest neighbor filling method is used for repair. Wherein, the KNN nearest neighbor filling method is preferably: querying the k non-null values data before and after the null value and filling the average value of the several non-null values data into the current null value.

[0089] When the degree of null values or duplicate values in the second modal data block determined in step 4-2 is large, the nearest neighbor linear interpolation method is used for repair. Wherein, the nearest neighbor linear interpolation method preferably performs linear interpolation by searching the non-null data before and after the null value time or stake position and the stake position relationship.​

[0090] Step 4-4, repeat steps 4-1 to 4-3 to sequentially complete the null value and repeated value repair of the third modal data block to the n-th modal data block.

[0091] In this case, the machine running modal data sampling frequency is high, the null value data is less and less than the set threshold, so it belongs to low null value degree modal data. The machine running modal data sampling frequency and the vibration frequency can be aligned at the second level, so the nearest neighbor KNN filling method is used for repair, and the specific implementation technology is to fill the null value in the machine running modal data with the mean value of its nearest neighbor k non-null values, and k=1 is selected in this case.

[0092] In this case, the cutter modal data sampling rate is low, the null value data is large and greater than the set threshold, so it belongs to high null value degree modal data. Therefore, the nearest neighbor linear interpolation method is used for repair. The specific implementation method is to query the wear value closest to the null value position time or stake number position, perform linear interpolation according to the non-null value and stake number spatial position, and fill the interpolation result to the null value position of the multi-modal fusion data.

[0093] For example, the cutter radial wear value in the stake number range of No. 32 tool position increases from 2 mm to 8 mm between 63926.3 and 63898.3, then the cutter wear rate in this stake number range can be calculated as (8 mm-2 mm) / (63926.3-63898.3)=0.2143 mm / m, and the cutter wear value in this stake number range can be calculated according to the wear rate and the corresponding stake number position by the interpolation formula, for example, the wear value at stake number position 63098.55 is 2+(63926.3-63098.55)×0.2143=5.80 mm.

[0094] Step 5, feature extraction: for each stability state type tunnel surrounding rock data filled n modal data blocks, each is taken as a unit for feature extraction; the same modal data block in the same excavation section is extracted according to the statistical or fitting method to obtain a plurality of characteristic values, so that n characteristic values corresponding to n modal data blocks in the same excavation section are obtained. sum = n1+n2+…+n n characteristic values; wherein the excavation section refers to the stage data in which the cutterhead speed value of the TBM decreases from a non-zero value to zero.

[0095] Step 6, constructing a TBM multi-modal feature database based on the excavation section: all the characteristic values of all the excavation sections extracted in step 5 of the four types of tunnel surrounding rock form a TBM multi-modal feature database based on the excavation section.

[0096] In this embodiment, the multi-modal data is extracted in units of excavation sections. Taking the three modal monitoring data on December 19, 2020 as an example, nine effective excavation sections are identified by the cutterhead speed reflecting the working state of the cutterhead. The features of different modal data within each excavation section are extracted. For vibration modal data, the maximum vibration amplitude, minimum vibration amplitude, and effective vibration amplitude of each excavation section are calculated. For excavation data, the excavation indicators of each excavation section are calculated, including the fitting slope a and intercept b of the thrust and penetration, the fitting slope TPI of the torque and penetration, and other parameters. For the cutter wear modal feature, the cutter wear speed parameters of 34 cutter positions are calculated according to the cutter wear value and the excavation distance. Finally, the cutterhead vibration, machine operation, and cutter wear feature database based on the excavation section is obtained as shown in Figure 6 .

[0097] In this embodiment, the TBM multi-modal feature database based on the excavation section includes the feature data of 5317 excavation sections.

[0098] Step 7, constructing a tunnel surrounding rock stability prediction model: the tunnel surrounding rock stability prediction model is a machine learning model, the input of the tunnel surrounding rock stability prediction model is n sum feature values of different modalities of an excavation section, and the output of the tunnel surrounding rock stability prediction model is the stability type of the tunnel surrounding rock. The machine learning model is preferably a prediction model based on gradient boosting ensemble learning algorithm.

[0099] Step 8, training the prediction model: using the TBM multi-modal feature database based on the excavation section constructed in step 6, the tunnel surrounding rock stability prediction model constructed in step 7 is trained.

[0100] First, the database is divided into training set and test set according to the four surrounding rock stability, and 4785 groups of data containing four surrounding rock stability states are used to construct the prediction model based on gradient boosting ensemble learning method. The test set prediction performance is shown in Figure 7 , the results show that the prediction accuracy and F1 score of the surrounding rock stability classification diagnosis using only machine operation modal data are 0.90 and 0.56 respectively, which indicates that many surrounding rock stability state categories are missed or misreported. When considering the machine excavation and cutter wear two modal data features, the accuracy and F1 score reach 0.97 and 0.92 respectively, which indicates that while the accuracy is significantly improved, the false positive rate and the false negative rate are decreased. However, when considering the machine excavation, cutter wear and cutterhead vibration three modal features, the accuracy of the surrounding rock stability reaches the highest, which indicates that the prediction method based on the fusion of multiple modalities can significantly improve the prediction performance. Finally, the surrounding rock stability prediction model trained based on multi-modal data is exported and deployed in the intelligent excavation platform of the host computer in the TBM control room.

[0101] Step 9, tunnel surrounding rock stability diagnosis: monitoring n kinds of modal data of TBM in the tunneling construction process, and obtaining n characteristic values of the current tunneling section; then, inputting the n characteristic values of the current tunneling section into the tunnel surrounding rock stability prediction model trained in step 8, and predicting the stability state of the current tunnel surrounding rock. sum sum sum

[0102] In the present example, the multi-modal data monitored during the excavation of the new work section are aligned and repaired, and the features are extracted in units of tunneling sections, so as to realize real-time diagnosis of the stability of the surrounding rock.

[0103] The present application realizes the standardization and normalization of the multi-modal data processing flow of independently collected different modal data in the construction site, further guarantees the accuracy of the alignment between different modal data based on time and space information, and realizes the repair and completion of modal data with low and high degrees of data null values through the nearest neighbor null value filling and linear interpolation method. The method is suitable for different monitoring modal fusion databases in the tunnel engineering construction process, reduces the time processing cost of different modal data, simultaneously realizes the rapid diagnosis of the stability of the surrounding rock on site, and guarantees the safety of the construction environment.

[0104] The above describes the preferred embodiments of the present application, but the present application is not limited to the specific details in the above embodiments, and various equivalent transformations can be made to the technical solutions of the present application within the technical concept of the present application, and these equivalent transformations all belong to the protection scope of the present application.​

Claims

1. A tunnel surrounding rock stability diagnosis method based on TBM multimodal data fusion, characterized by: The steps include: Step 1: Classify surrounding rock stability types: Classify surrounding rock stability into N types of stability states; where N ≥ 2; Step 2: TBM multimodal data preparation: For each stability state type of tunnel surrounding rock, TBM excavation is carried out. During the tunnel excavation process, different types of modal data are continuously monitored and collected. As a result, n types of modal data are collected for each stability state type of tunnel surrounding rock. Subsequently, each modal data for each stability state type of tunnel surrounding rock is segmented into n modal data blocks on a daily basis. Step 3: Temporal and spatial alignment of TBM multimodal data: For each stability state type, the n modal data blocks of the tunnel surrounding rock are sorted from high to low according to sampling frequency. Taking the modal data block corresponding to the highest sampling frequency as the benchmark, the remaining n-1 modal data blocks are traversed based on the time and space stake positions and copied to the modal data block corresponding to the highest sampling frequency for alignment and fusion processing. Step 4: Data filling: Identify null and duplicate values ​​in the spatiotemporally aligned TBM multimodal data in step 4. Repair the identified null and duplicate values ​​using the KNN nearest neighbor filling or linear interpolation method by determining the percentage of null and duplicate values. Step 5, feature extraction: for each stability state type, the n modal data blocks filled with tunnel surrounding rock data are subjected to feature extraction in units of excavation sections. For the same modal data block in the same excavation section, several values ​​are extracted as feature values ​​by statistical or physical index calculation methods. Therefore, n modal values ​​can be obtained for the i-th mode in the same excavation section. i features, then for all n modal data blocks corresponding to n sum =n1+n2+…+n n The driving section refers to the stage data when the cutter head speed of the TBM changes from zero to non-zero and finally drops to zero during the process from the start of the cutter head to the stop at the maximum stroke; Step 6: Construct a TBM multimodal feature database based on the excavation section: Based on the characteristic values ​​of the different modal data of all excavation sections extracted for different surrounding rock stability working conditions in step 5, a TBM multimodal feature database based on the excavation section is constructed; Step 7: Construct a tunnel surrounding rock stability prediction model. The tunnel surrounding rock stability prediction model is a machine learning model. The input of the tunnel surrounding rock stability prediction model is the characteristic values ​​of n modes within a tunneling section, and the output of the tunnel surrounding rock stability prediction model is the tunnel surrounding rock stability category. The tunnel surrounding rock stability prediction model is trained using the tunneling section-based TBM multimodal feature database constructed in Step 6. Step 8: Tunnel surrounding rock stability diagnosis: Monitor n types of modal data during the tunnel excavation process of the TBM and obtain a total of n different modal data for the current excavation section. sum Then, the n eigenvalues ​​of the current excavation section are sum The characteristic values ​​are input into the tunnel surrounding rock stability prediction model trained in step 7 to predict the stability state of the tunnel surrounding rock in the current excavation section.

2. The tunnel surrounding rock stability diagnosis method based on TBM multimodal data fusion according to claim 1 is characterized by: In step 2, the modal data includes TBM operation data, cutterhead cutter wear data, and cutterhead vibration data.

3. The tunnel surrounding rock stability diagnosis method based on TBM multimodal data fusion according to claim 2 is characterized by: In step 2, the TBM operation data includes thrust, torque, propulsion speed, rotational speed, and penetration; the cutterhead cutter wear data includes the radial wear data of the cutterhead at each cutter position on the cutterhead; and the cutterhead vibration data includes the vibration acceleration data at the cutterhead monitoring position.

4. The tunnel surrounding rock stability diagnosis method based on TBM multimodal data fusion according to claim 1 is characterized by: In step 3, the spatiotemporal alignment method of TBM multimodal data includes the following steps: Step 3-1, inter-modal sorting: sort the n modal data blocks after data segmentation from high to low according to the sampling frequency, namely: first modal data block, second modal data block, ..., nth modal data block; Step 3-2, Intra-modal Sorting: Sort each modal data block in the order of monitoring time or spatial stake position; Step 3-3, Alignment of Second Modal Data Block: Based on the time and spatial stake position, taking the time information of the first modal data block as the reference, and through the time index, insert the data at each acquisition frequency in the second modal data block into the first modal data block with the same time accuracy to achieve data alignment and fusion of the second modal data block and the first modal data block; Step 3-4: Repeat step 3-3 until the data alignment and fusion of the nth modal data block and the first modal data block are completed.

5. The tunnel surrounding rock stability diagnosis method based on TBM multimodal data fusion according to claim 4 is characterized in that: When filling the data in step 4, it is necessary to first determine the degree of null values ​​and duplicate values ​​in the modal data; for modal data blocks with a small degree of null values ​​and duplicate values, the KNN nearest neighbor filling method is used to repair the null values ​​and duplicate values; for modal data blocks with a large degree of null values ​​and duplicate values, the nearest neighbor linear interpolation method is used to repair the null values ​​and duplicate values.

6. The tunnel surrounding rock stability diagnosis method based on TBM multimodal data fusion according to claim 5 is characterized by: In step 4, the specific method of filling data includes the following steps: Step 4-1, identifying null values ​​and duplicate values: identifying null values ​​and duplicate values ​​in the aligned and fused second modality data block; Step 4-2, determining the degree of null values ​​and duplicate values: comparing the number of null values ​​and duplicate values ​​of each modality in each day with the set ratio value. When the number of null values ​​and duplicate values ​​in the i-th day is less than the set ratio value, all null values ​​and duplicate values ​​in the i-th day are judged as null values ​​or with a low degree of duplication; otherwise, they are judged as null values ​​or with a high degree of duplication; Step 4-3, repairing null values ​​and duplicate values: When the degree of null values ​​or duplicate values ​​in the second modal data block determined in step 4-2 is small, the KNN nearest neighbor filling method is used for repair; when the degree of null values ​​or duplicate values ​​in the second modal data block determined in step 4-2 is large, the nearest neighbor linear interpolation method is used for repair; Step 4-4: Repeat steps 4-1 to 4-3 to complete the repair of null values ​​and duplicate values ​​of the third modal data block to the nth modal data block in sequence.

7. The tunnel surrounding rock stability diagnosis method based on TBM multimodal data fusion according to claim 1 is characterized by: In step 5, feature extraction is performed on different modal data within the same excavation section. For the machine operation mode, statistical characteristics and linear fitting methods are used to extract the mean and rock excavation indicators; for the wear mode, the radial wear velocity values ​​of the cutter at different cutter positions are calculated as characteristic parameters; for the cutterhead vibration mode, the statistical characteristic values ​​of the vibration acceleration are extracted as characteristic parameters.

8. The tunnel surrounding rock stability diagnosis method based on TBM multimodal data fusion according to claim 1 is characterized by: In step 7, the deep learning model is an evaluation model based on the gradient-based boosting ensemble learning algorithm.

9. The tunnel surrounding rock stability diagnosis method based on TBM multimodal data fusion according to claim 1 is characterized by: In step 1, N=5, which are stable, basically stable, partially stable, unstable and extremely unstable.

10. The tunnel surrounding rock stability diagnosis method based on TBM multimodal data fusion according to claim 1, characterized in that: In step 4-2, the ratio is set to 5% or 10% of the total number of the corresponding modalities in the corresponding day.

Images