Prediction method of shield tunneling machine cutterhead torque parameter

Abstract

The invention relates to a shield tunneling machine cutterhead torque parameter prediction method. The method comprises the steps that equipment operation data and geological environment data in the tunneling process of a shield tunneling machine are periodically collected; performing data preprocessing on the equipment operation data and the geological environment data to obtain preprocessed data; determining a stability data sequence in the preprocessed data and a difference order corresponding to the stability data sequence through ADF stability test, and obtaining parameters of stability data sequence adaptive optimization based on a Bayesian information criterion to generate an optimal ARIMA model; performing prediction processing on the preprocessed data by adopting an optimal ARIMA model to obtain a predicted value of the shield tunneling machine cutterhead torque parameter; adjusting the tunneling speed of the shield tunneling machine and the rotating speed of the cutter head in the next period according to the predicted value of the torque parameter of the cutter head; and the steps are repeated until the shield tunneling machine is turned off. The construction cost of the shield tunneling machine is reduced, and the engineering efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of shield machine construction, and particularly relates to a method for predicting the cutter head torque parameters of a shield machine. Background Art

[0002] With the rapid development of urban underground space development, the shield method has become the core construction method for underground projects such as subway tunnels and utility tunnels due to its advantages of high construction efficiency and little environmental impact. As the key equipment for shield method construction, the accurate prediction of the cutter head torque parameters of a shield machine is directly related to the tunneling efficiency, equipment safety, and engineering cost control. However, during shield construction, complex and variable geological conditions are faced, resulting in non-stationarity and mutation characteristics of the cutter head torque, making it insufficiently adaptable in complex strata construction and seriously affecting the reliability of the shield method implementation.

[0003] Therefore, there is an urgent need for a method for predicting the cutter head torque parameters of a shield machine. Summary of the Invention

[0004] (I) Technical Problems to be Solved In view of the above-mentioned disadvantages and deficiencies of the prior art, the present invention provides a method for predicting the cutter head torque parameters of a shield machine, which solves the technical problems in the prior art of being unable to adapt to the time-varying characteristics of geology, difficult to meet the real-time control requirements, and prediction inaccuracy caused by engineering data interference.

[0005] (II) Technical Solutions In order to achieve the above object, the main technical solutions adopted by the present invention include: An embodiment of the present invention provides a method for predicting the cutter head torque parameters of a shield machine, including: S100. Periodically collect the equipment operation data and geological environment data during the tunneling process of the shield machine; S200. Perform data preprocessing on the equipment operation data and geological environment data to obtain the preprocessed data; S300. Determine the stable data sequence and the corresponding difference order in the preprocessed data through the ADF stationarity test method; S400. Obtain the parameters for adaptive optimization of the stable data sequence based on the Bayesian information criterion, and generate an optimal ARIMA model for predicting the cutter head torque parameters of the shield machine; S500. Use the optimal ARIMA model to perform prediction processing on the preprocessed data to obtain the predicted value of the cutter head torque parameters of the shield machine; S600. Adjust the tunneling speed and cutter head rotation speed of the shield machine in the next cycle according to the predicted value of the cutter head torque parameters.

[0006] Optionally, in the S100, The device operation data includes cutter head torque, tunneling speed, cutter head rotation speed, and soil chamber pressure; the geological environment data includes physical and mechanical parameters of rock and soil masses, geological structure characteristic data, and hydrogeological parameter data.

[0007] Optionally, in S200, the data preprocessing of the device operation data and the geological environment data includes: S210. Perform sliding window dynamic management on the device operation data and the geological environment data to obtain the data of the current window, and the window length includes 100 - 2000 data points; S220. Remove the data of the empty - push section from the device operation data and the geological environment data of the current cycle to obtain the data after removing the empty - push section data; the data of the empty - push section is the data where the cutter head rotation speed is greater than zero and the tunneling speed is less than 5 mm / min; S230. Filter the outliers in the data after removing the empty - push section data by using the 3σ criterion to obtain the data after filtering the outliers; S240. Extract the trend term in the data after filtering the outliers through singular spectrum analysis to obtain the data after removing the trend term as the preprocessed data.

[0008] Optionally, S300 specifically includes: Perform ADF test on the preprocessed data to obtain the P - value and the data sequence corresponding to the P - value; When the P - value of the test of the preprocessed data ≤ 0.05, take the optimal difference order as 0, and obtain the data sequence at this time as the stable data sequence; when the P - value of the ADF test of the sequence after the first - order difference ≤ 0.05, take the optimal difference order as 1, and obtain the data sequence at this time as the stable data sequence; when the P - value of the ADF test of the sequence after N - order difference ≤ 0.05, take the optimal difference order as N, and obtain the data sequence at this time as the stable data sequence; N is a natural number.

[0009] Optionally, S400 includes: S410. Based on the Bayesian information criterion, traverse the combinations of autoregressive orders and moving average orders within a preset range, and select the combination of autoregressive order and moving average order when the Bayesian information criterion value is the smallest in the stable data sequence; S420. Obtain the optimal ARIMA model according to the difference order, autoregressive order, and moving average order combination corresponding to the stable data sequence.

[0010] Optionally, S410 includes: S411. Within the preset range of autoregressive order ∈ [0, 5] and moving average order ∈ [0, 5], traverse all combinations of autoregressive order and moving average order based on the Bayesian Information Criterion (BIC) to obtain the BIC value for each combination. The BIC value is obtained through the following formula: ; where BIC is the Bayesian Information Criterion value, L is the model likelihood function, representing the goodness of fit to the stability data sequence, n is the number of samples, p is the autoregressive order, and q is the moving average order. S412. Select the combination of autoregressive order and moving average order with the smallest BIC value as the optimal order combination.

[0011] Optionally, S410 specifically includes: S413. Divide the preset range of autoregressive order and moving average order into k subtasks, where k is the number of computing nodes. S414. Assign each subtask to an independent computing node through the CUDA framework based on GPU, and each node calculates the BIC value in parallel based on the assigned order combination range. S415. Use a message queue to collect the BIC values of all computing nodes in real time, and extract the combination of autoregressive order and moving average order corresponding to the minimum BIC value through a reduction operation.

[0012] Optionally, the method further includes: In a formation with uneven hardness, if the optimal differential order ≥ 2, synchronously increase the sliding window length to 1.5 times the original window.

[0013] Optionally, S600 specifically includes: S610. When the predicted value of the cutterhead torque parameter exceeds 20% of the theoretical torque value under the current geological conditions, reduce the cutterhead rotation speed by 0.5 - 1.0 r / min and reduce the tunneling speed to 80% - 90% of the original set value. S620. When the predicted value of the cutterhead torque parameter is within the range of ±15% of the theoretical torque value under the current geological conditions, increase the tunneling speed to 105% - 110% of the original set value.

[0014] Optionally, S230 includes: S231. Obtain the mean μ and standard deviation σ of the data after removing the data in the empty push section; set the outlier judgment threshold as the interval range from μ - 3σ to μ + 3σ. S232. Traverse and remove the data of the empty push section. If any parameter value of the cutterhead torque, tunneling speed, or soil chamber pressure exceeds the corresponding threshold range, it is determined as an outlier and removed, and the preprocessed data is obtained.

[0015] (III) Beneficial effects The beneficial effects of the present invention are as follows: For a prediction method of the cutterhead torque parameter of a shield machine according to the present invention, since the ARIMA model with adaptive order optimization is adopted, the ARIMA model can continuously change the model optimization parameters according to data updates, which well solves the problem of difficultly capturing the data change law under the conditions of a new stratum when the stratum changes. At the same time, from data acquisition to parameter adjustment, the present invention truly realizes prediction and control, which not only improves the accuracy and precision of prediction, but also enhances the real-time performance and robustness of prediction, and realizes the improvement of engineering efficiency. Description of the drawings

[0016] Figure 1 It is a schematic flow chart of a prediction method of the cutterhead torque parameter of a shield machine according to Embodiment 1 of the present invention; Figure 2 It is a flow chart for optimizing the parameters of the adaptive ARIMA model in the embodiments of the present invention. Specific implementation manners

[0017] In order to better explain the present invention for easy understanding, the present invention will be described in detail below with reference to the drawings through specific implementation manners.

[0018] During shield tunneling, complex geological conditions such as uneven hard and soft strata, fluctuating groundwater levels, and developed joint fissures are often faced, resulting in non-stationarity and mutation characteristics of the cutterhead torque. Existing methods rely on the ARIMA model with fixed parameters or empirical formulas and cannot dynamically adjust the model parameters according to stratum changes, resulting in prediction lags. At the same time, the prediction errors of existing methods are generally > 30%, resulting in low prediction accuracy and affecting tunneling efficiency and equipment safety.

[0019] The embodiments of the present invention can be applied to the construction of urban subway tunnels. In the stratum where soft soil and hard rock alternate, by predicting the change of the cutterhead torque parameter, the tunneling parameters are dynamically adjusted to avoid cutterhead jamming. The embodiments of the present invention can also be applied to the environment with a high groundwater level and fluctuating permeability coefficient. Combining hydrogeological parameters, the change of the cutterhead torque parameter is predicted and relevant parameters are adjusted. Further, during the tunneling of mine roadways, the broken stratum with developed joint fissures can be dealt with. By predicting the cutterhead torque in real time, the cutterhead rotation speed is adjusted to reduce tool wear.

[0020] The present invention solves the problems of low prediction accuracy and poor real-time performance of the cutterhead torque in shield tunneling through the adaptive ARIMA model and multi-source data fusion technology, and provides reliable technical support for intelligent construction under complex geological conditions.

[0021] To better understand the above technical solution, the exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a clearer and more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0022] Embodiment 1

[0023] See Figure 1 , a prediction method for the cutter head torque parameters of a shield machine according to an embodiment of the present invention includes: Step S100, periodically collect the equipment operation data and geological environment data during the tunneling process of the shield machine; Step S200, perform data preprocessing on the equipment operation data and geological environment data to obtain the preprocessed data; Step S300, determine the stable data sequence and the corresponding difference order of the stable data sequence in the preprocessed data through the ADF stationarity test method; Step S400, obtain the parameters for adaptive optimization of the stable data sequence based on the Bayesian information criterion, and generate an optimal ARIMA model for predicting the cutter head torque parameters of the shield machine; Step S500, perform prediction processing on the preprocessed data using the optimal ARIMA model to obtain the predicted values of the cutter head torque parameters of the shield machine; Step S600, adjust the tunneling speed and cutter head rotation speed of the shield machine in the next cycle according to the predicted values of the cutter head torque parameters.

[0024] This embodiment further includes: Step S700, repeatedly execute steps S100 to S600 until the shield machine shuts down.

[0025] After the tunneling parameters are updated, the window slides forward to obtain new equipment operation data and geological environment data, and recalculate the optimal model parameters of the current sequence. Thus, the adaptive determination of the optimal parameters of the ARIMA prediction model can be completed, so that the prediction accuracy of the prediction method for the cutter head torque parameters of the shield machine of the present invention always remains at a high level.

[0026] In this embodiment, the equipment operation data includes cutter head torque, tunneling speed, cutter head rotation speed, and soil chamber pressure; the geological environment data includes geotechnical physical and mechanical parameters, geological structure characteristic data, and hydrogeological parameter data.

[0027] Among them, the equipment operation data refers to the real-time state parameters of the equipment directly measured by the sensors on the shield machine body. The cutter head torque is measured by a torque sensor, the tunneling speed is collected by a laser displacement sensor, the soil chamber pressure is obtained through a pressure sensor array with 6 - 8 measuring points, and the cutter head rotation speed is measured by an optoelectronic encoder.

[0028] The geological environment data refers to the formation parameters in front of the tunneling obtained through external detection means. Among them, the physical and mechanical parameters of the rock and soil mass include the compression modulus inverted from the reflected waves of the TSP geological radar, the cohesion and internal friction angle obtained based on the gray-scale analysis of the muck image of the screw conveyor, the geological structure characteristic data includes the dip angle of the formation interface and the degree of joint fissure development, and the hydrogeological parameter data includes the groundwater level height and the permeability coefficient.

[0029] In this embodiment, through the multi-source fusion of the equipment operation data and the geological environment data, the subsequent optimization of the ARIMA model is carried out, forming a multi-scale prediction framework. Especially in the hard and soft interaction formation, the prediction error is significantly reduced.

[0030] A prediction method for the cutter head torque parameter of a shield machine in this embodiment provides a basis for the optimization, prediction and control of shield tunneling parameters and the informatization construction of shield tunnels, reduces the construction cost of the shield machine, and can improve the engineering efficiency.

[0031] To better understand the foregoing method, this embodiment gives a prediction method for the cutter head torque parameter of a shield machine in combination with a specific process, including: Step S100: Periodically collect the equipment operation data and the geological environment data during the tunneling of the shield machine; Step S200: Perform data preprocessing on the equipment operation data and the geological environment data to obtain the preprocessed data; Step S300: Determine the stable data sequence and the corresponding difference order of the stable data sequence in the preprocessed data through the ADF stationarity test method; Step S400: Obtain the parameters for adaptive optimization of the stable data sequence based on the Bayesian information criterion, and generate the optimal ARIMA model for predicting the cutter head torque parameter of the shield machine; Step S500: Use the optimal ARIMA model to perform prediction processing on the preprocessed data to obtain the predicted value of the cutter head torque parameter of the shield machine; Step S600: According to the predicted value of the cutter head torque parameter, adjust the tunneling speed and the cutter head rotation speed of the shield machine in the next cycle.

[0032] Step S700: Repeat steps S100 to S600 until the shield machine shuts down.

[0033] Among them, the equipment operation data includes cutter head torque, tunneling speed, cutter head rotation speed, and soil chamber pressure; the geological environment data includes physical and mechanical parameters of rock and soil masses, geological structure characteristic data, and hydrogeological parameter data.

[0034] In step S200, the data preprocessing of the equipment operation data and the geological environment data includes: Step S210: Perform sliding window dynamic management on the equipment operation data and the geological environment data to obtain the data in the current window, and the window length includes 100 - 2000 data points; Step S220: Remove the data in the empty - push section from the equipment operation data and the geological environment data in the current cycle to obtain the data after removing the empty - push section data; the empty - push section data is the data with the cutter head rotation speed greater than zero and the tunneling speed less than 5 mm / min; Step S230: Filter the outliers in the data after removing the empty - push section data by using the 3σ criterion to obtain the data after filtering out the outliers; Step S230 includes: Step S231: Obtain the mean μ and standard deviation σ of the data after removing the empty - push section data; set the outlier judgment threshold as the interval range from μ - 3σ to μ + 3σ; Step S232: Traverse the data after removing the empty - push section data. If any parameter value of the cutter head torque, tunneling speed, or soil chamber pressure exceeds the corresponding threshold interval, it is determined as an outlier and removed to obtain the pre - processed data.

[0035] Step S240: Extract the trend term in the data after filtering out the outliers through singular spectrum analysis, and obtain the data after removing the trend term as the pre - processed data.

[0036] The specific steps of step S240 are: Construct a trajectory matrix: Arrange the pre - processed data according to the window length to generate a trajectory matrix with dimensions of L×K, where, , K = N w-L+1 , N w is the sliding window length, and w is the w - th window; Singular value decomposition: Perform SVD decomposition on the trajectory matrix to obtain the eigenvalues and the corresponding eigenvectors; Trend term reconstruction: Group the components according to the singular value magnitude and physical meaning; Select the components corresponding to the first k main singular values, usually the first 1 - 2 principal components, and reconstruct the trend sub - matrix with the corresponding left singular vector and right singular vector. The remaining components are classified as periodic or noise terms.

[0037] Perform diagonal averaging on the trend sub - matrix to convert it into a time - series - form trend term. Subtract the trend term from the original data to generate the pre - processed data.

[0038] Further, step S300 specifically includes: Perform an ADF test on the preprocessed data to obtain the P-value and the data sequence corresponding to the P-value; When the P-value of the preprocessed data test ≤ 0.05, take the optimal difference order as 0, and obtain the data sequence at this time as the stable data sequence; when the P-value of the ADF test of the sequence after the first difference ≤ 0.05, take the optimal difference order as 1, and obtain the data sequence at this time as the stable data sequence; when the P-value of the ADF test of the sequence after N differences ≤ 0.05, take the optimal difference order as N, and obtain the data sequence at this time as the stable data sequence; N is a natural number.

[0039] The ADF test is an important method in time series analysis for judging whether a sequence is stationary. Its core is to determine stationarity by testing whether there is a unit root in the sequence. Specifically, assume that H0 is that the sequence has a unit root, indicating non-stationarity, and assume that H1 is that the sequence does not have a unit root, indicating stationarity. When testing, it is necessary to select an appropriate model form according to the characteristics of the sequence. The model forms include no constant term and trend term, containing a constant term, containing a constant term and trend term, and determine the lag order through the information criterion and calculate the corresponding P-value. If the P-value ≤ 0.05, then reject the original hypothesis and determine that the sequence is stationary; if the P-value > 0.05, then accept the original hypothesis and determine that the sequence is non-stationary, and subsequent processing is required.

[0040] When the ADF test result is non-stationary, it is necessary to perform difference processing on the sequence to eliminate the trend or seasonal influence: First, perform a first-order difference on the non-stationary sequence, that is, subtract the previous value from the current value to obtain a new sequence after the first-order difference. Subsequently, perform an ADF test on this new sequence again to determine whether it is stationary. If the sequence after the first-order difference is still non-stationary, continue to perform a second-order difference and repeat the ADF test until a stationary sequence is obtained.

[0041] In the specific implementation process, the maximum order of the above difference processing is 3; when the formation change causes the data to be non-stationary, such as transitioning from soft soil to hard rock, the difference order is automatically increased by 1 or 2 to eliminate the trend and seasonal influence and ensure the stationarity of the data.

[0042] In this embodiment, step S400 includes: Step S410, based on the Bayesian information criterion, traverse the combination of autoregressive order and moving average order within the preset range, and select the combination of autoregressive order and moving average order when the Bayesian information criterion value is the smallest in the stable data sequence; Step S420: Obtain the optimal ARIMA model according to the combination of the difference order, autoregressive order, and moving average order corresponding to the stability data sequence.

[0043] Step S410 includes: Step S411: Within the preset range where the autoregressive order ∈ [0, 5] and the moving average order ∈ [0, 5], traverse all combinations of autoregressive orders and moving average orders based on the Bayesian information criterion, and obtain the Bayesian information criterion value for each combination. The Bayesian information criterion value is obtained through the following formula: ; where BIC is the Bayesian information criterion value, L is the model likelihood function, representing the goodness of fit of the stability data sequence, n is the number of samples, p is the autoregressive order, and q is the moving average order. S412: Select the combination of autoregressive order and moving average order with the smallest Bayesian information criterion value as the optimal order combination.

[0044] In a specific implementation process, Step S410 specifically includes: Step S413: Divide the preset range of autoregressive order and moving average order into k subtasks, where k is the number of computing nodes; specifically, divide the preset range of autoregressive order p ∈ [0, 5] and moving average order q ∈ [0, 5] into 6×6 = 36 discrete points, and each point represents a (p, q) combination; allocate tasks according to the number of computing nodes k. If k = 4, then each node is assigned 9 combinations. For example, node 1 processes p = 0 - 2, q = 0 - 5. Adopt the main node dynamic allocation mechanism. The main node maintains a task queue and initially pushes all (p, q) combinations into the queue; after a computing node completes the current task, it requests a new task from the main node through the gRPC protocol; at the same time, set a task timeout threshold, such as 5 seconds, and timeout tasks are automatically reallocated to other nodes.

[0045] Step S414: Allocate each subtask to an independent computing node through the CUDA framework based on GPU, and each node calculates the Bayesian information criterion value in parallel based on the allocated order combination range. Each computing stage is responsible for calculating the Bayesian information criterion value of a (p, q), loading the ARIMA model parameters corresponding to the order from the global memory; calling the CUDA Math API to calculate the log-likelihood function ln(L) in parallel; and outputting the result according to the formula for calculating the Bayesian information criterion value above. In addition, when the execution fails, trigger exception capture and restart the corresponding computing node; for NAN or lnf outlier values, automatically mark the corresponding (p, q) as invalid. Step S415: Use a message queue to collect the Bayesian Information Criterion (BIC) values of all computing nodes in real time, and extract the autoregressive order and moving average order combination corresponding to the minimum BIC value through a reduction operation.

[0046] Among them, step S415 can be specifically divided into: Message queue architecture: RabbitMQ is used to implement distributed node communication. Each computing node encapsulates the BIC value and the corresponding (p, q) combination into a JSON message. Implementation of reduction operation: Inside each computing node, CUDA atomic operations are used to find the minimum BIC value in the current task. After the master node receives all local minima, the global optimal (p, q) combination is determined through a red-black tree sorting algorithm. Guarantee of real-time performance: Set the maximum delay threshold of the message queue, such as 500 ms. Packets that do not arrive within the timeout will be discarded and trigger local data retransmission. At the same time, a priority queue mechanism is adopted to assign higher processing priority to low-order combinations with p + q ≤ 3.

[0047] Through the above parallel computing, the single-parameter traversal time is controlled within 1.5 seconds, improving the running speed.

[0048] A prediction method for the cutter head torque parameters of a shield machine in this embodiment further includes: In a formation with uneven hardness, if the optimal difference order ≥ 2, the sliding window length is synchronously increased to 1.5 times the original window.

[0049] Furthermore, S600 in this embodiment specifically includes: Step S610: When the predicted value of the cutter head torque parameter exceeds 20% of the theoretical torque value under the current geological conditions, lower the cutter head rotation speed by 0.5 - 1.0 r / min and reduce the tunneling speed to 80% - 90% of the original set value. Among them, the theoretical torque value is calculated through the earth pressure balance formula, specifically: ; where M is the formation friction coefficient, σ v is the vertical ground stress, D is the cutter head diameter, and L is the tunneling length.

[0050] Step S620: When the predicted value of the cutter head torque parameter is within the range of ±15% of the theoretical torque value under the current geological conditions, increase the tunneling speed to 105% - 110% of the original set value.

[0051] In this embodiment, the predicted value of the cutter head torque parameter realizes the optimization of tunneling parameters through the full-link mechanism of predicted value real-time feedback - control strategy mapping - parameter dynamic adjustment.

[0052] First, the optimal ARIMA model outputs the predicted value of the cutter head torque parameter, which is transmitted to the shield machine PLC controller in real time through the industrial communication protocol. Together with the actual torque, tunneling speed, cutter head rotation speed, soil chamber pressure, and geological environment data collected in real time, they constitute the input of the control layer. The control layer combines PID control and feedforward control to establish the mapping relationship between the predicted value of the cutter head torque parameter and the tunneling parameters: when the predicted value of the cutter head torque parameter exceeds 20% of the theoretical torque value under the current geological conditions, the cutter head rotation speed is reduced by 0.5 - 1.0 r / min, and the tunneling speed is reduced to 80% - 90% of the original set value; when the predicted value of the cutter head torque parameter is within the range of ±15% of the theoretical torque value under the current geological conditions, the tunneling speed is increased to 105% - 110% of the original set value.

[0053] A prediction method for the cutter head torque parameter of a shield machine in this embodiment further includes a formation mutation detection mechanism. When the torque prediction error exceeds 15% within three consecutive prediction windows, model reconstruction is triggered, and the sliding window is automatically reduced to 50% of the original length for local parameter refitting.

[0054] A prediction method for the cutter head torque parameter of a shield machine in this embodiment is deployed on the edge computing gateway of the shield machine, and the specific configuration is as follows: An industrial-grade NVIDIA Jetson Xavier module is adopted, with a built-in TensorRT acceleration engine; it communicates with the shield machine PLC controller through the OPC UA protocol, and the end-to-end delay < 300 ms; a built-in abnormal fuse mechanism is provided, and when the CPU utilization rate > 90%, it automatically switches to the degraded mode and only outputs the safety threshold of the propulsion force.

[0055] A prediction method for the cutter head torque parameter of a shield machine in this embodiment adopts a recursive differential order adaptive and preset range joint optimization mechanism, which can dynamically adjust the ARIMA model structure in real time, enabling the ARIMA model to continuously replace the optimal parameters of the model according to data updates, and well solving the problem that it is difficult for the prediction model to capture the data change law under the new formation conditions when the formation changes, while improving the prediction accuracy. In addition, the present invention forms a millisecond-level response link from data collection to parameter adjustment, truly realizing "prediction is control", and providing a replicable algorithm engineering paradigm for the intelligentization of shield machines.

[0056] A prediction method for the cutter head torque parameter of a shield machine in this embodiment specifically includes: Step S100: Periodically collect the equipment operation data and geological environment data during the tunneling process of the shield machine; Step S200: Perform data preprocessing on the equipment operation data and geological environment data to obtain the preprocessed data; Step S300: Determine the stable data sequence and the corresponding difference order in the preprocessed data through the ADF stationarity test method; Step S400: Obtain the parameters for adaptive optimization of the stable data sequence based on the Bayesian information criterion, and generate the optimal ARIMA model for predicting the cutterhead torque parameters of the shield machine; Step S500: Use the optimal ARIMA model to perform prediction processing on the preprocessed data to obtain the predicted values of the cutterhead torque parameters of the shield machine; Step S600: Adjust the tunneling speed and cutterhead rotation speed of the shield machine in the next cycle according to the predicted values of the cutterhead torque parameters.

[0057] Step S700: Repeat Steps S100 to S600 until the shield machine shuts down.

[0058] See Figure 2 , and the specific steps of Step S300 and Step S400 are as follows: First, to reduce the iteration time of each calculation, set a fixed window of length N w in the algorithm. Then, under the condition of continuous data update, only use the data in the first N w points at the current moment to establish the ARMIA model. After that, it is necessary to perform stationarity analysis on the data in the window. Furthermore, perform stationarity analysis on the time series, construct a test statistic to judge the stability of the time series, and the main test method is the ADF test. Determine whether the time series is stationary by determining the existence of unit roots in the sequence. Generally, select a confidence level of 0.05. If the P-value obtained from the ADF test on the time series is greater than 0.05, the sequence is non-stationary. While using the ADF test to perform stationarity test on the sequence, determine the optimal d value (optimal difference order). The specific steps to find the best d value are as follows: Based on the first N w sequences at the current sampling moment k: ; x k is the sequence at the kth sampling moment, N w is the length of the basic sliding window, N w+1 , N w+2 represent the time series offsets of each data point in the window. For example, if N w = 100, the window ranges from k - 99 to k, a total of 100 points. Let be the optimal estimate of the current d value of ARIMA. Perform an ADF test on the sequence . If the P-value is less than 0.05, it means the sequence is stationary, and take = 0. If the P - value is greater than 0.05, the sequence is differenced once. If the P - value calculated after one - order differencing is less than 0.05, take = 1; Similarly, for = 2, = 3, etc. are also obtained through the above - mentioned processing. After the above - mentioned processing, the best estimate of the d value can be completed, and at the same time, the sequence after times of differencing is obtained .

[0059] After determining the optimal differencing order, the Bayesian information criterion is used to optimize p and q for the differenced sequence to determine the optimal order. The specific optimization steps are as follows: First, assume that the best model orders p and q are respectively in the value ranges of

p1, p m

q1, q m

p1, p m

q1, q m

[0060] After the optimization process of the three parameters p, q, and d, the three optimal parameters of the ARIMA model for the sequence within the window with the current window length of N w can be determined, and the predicted value of the current sequence is output. When the input data is updated, the window slides forward to obtain a new engineering data sequence, and the optimal model parameters of the current sequence are recalculated. Thus, the adaptive determination of the optimal parameters of the ARIMA prediction model can be completed, making the prediction accuracy of a prediction method for the cutterhead torque parameters of a shield machine of the present invention always remain at a relatively high level.

[0061] A prediction method for the cutter head torque parameters of a shield machine in this embodiment incorporates an adaptive order optimization algorithm on the basis of the ARIMA model, enabling the ARIMA model to continuously replace the optimal parameters of the model according to data updates, solving the problems of real-time performance and robustness of the traditional ARIMA model under complex geological conditions, and improving the accuracy and efficiency of cutter head torque parameter prediction.

[0062] In the description of the present invention, it should be understood that the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, "a plurality of" means two or more unless otherwise specifically defined.

[0063] In the present invention, unless otherwise clearly specified and defined, terms such as "installation", "connection", "connection", "fixation" and the like should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral body; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium; it may be the communication inside two components or the interaction relationship between two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0064] In the present invention, unless otherwise clearly specified and defined, the first feature being "on" or "under" the second feature may be that the first and second features are in direct contact, or the first and second features are indirectly in contact through an intermediate medium. Moreover, the first feature being "above", "over" and "on" the second feature may be that the first feature is directly above or obliquely above the second feature, or merely indicates that the first feature has a higher horizontal height than the second feature. The first feature being "under", "below" and "beneath" the second feature may be that the first feature is directly below or obliquely below the second feature, or merely indicates that the first feature has a lower horizontal height than the second feature.

[0065] In the description of this specification, the descriptions of terms such as "one embodiment", "some embodiments", "embodiment", "example", "specific example", or "some examples" mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0066] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for predicting the torque parameter of a shield machine cutter head, characterized in that Including: S100. Periodically collect the equipment operation data and geological environment data during the tunneling process of the shield machine; S200. Perform data preprocessing on the equipment operation data and geological environment data to obtain the preprocessed data; S300. Determine the stable data sequence and the corresponding difference order in the preprocessed data through the ADF stationarity test method; S400. Obtain the parameters for adaptive optimization of the stable data sequence based on the Bayesian information criterion, and generate the optimal ARIMA model for predicting the cutter head torque parameters of the shield machine; S500. Use the optimal ARIMA model to perform prediction processing on the preprocessed data to obtain the predicted values of the cutter head torque parameters of the shield machine; S600. Adjust the tunneling speed and cutter head rotation speed of the shield machine in the next cycle according to the predicted values of the cutter head torque parameters.

2. The prediction method of the cutter head torque parameter of the shield machine according to claim 1, characterized in that In the S100, the equipment operation data includes cutter head torque, tunneling speed, cutter head rotation speed, and chamber pressure; the geological environment data includes geotechnical physical and mechanical parameters, geological structure characteristic data, and hydrogeological parameter data.

3. The prediction method of the cutter head torque parameter of the shield machine according to claim 1, characterized in that In the S200, the data preprocessing of the equipment operation data and geological environment data includes: S210. Perform sliding window dynamic management on the equipment operation data and geological environment data to obtain the data in the current window, and the window length includes 100 - 2000 data points; S220. Remove the data in the empty - push section from the equipment operation data and geological environment data in the current cycle to obtain the data after removing the empty - push section data; the empty - push section data is the data with the cutter head rotation speed greater than zero and the tunneling speed less than 5 mm / min; S230. Filter the outliers in the data after removing the empty - push section data using the 3σ criterion to obtain the data after filtering out the outliers; S240. Extract the trend term in the data after filtering out the outliers through singular spectrum analysis, and obtain the data after removing the trend term as the preprocessed data.

4. The prediction method for the cutter head torque parameters of a shield machine according to claim 1, wherein, The S300 specifically includes: Perform the ADF test on the preprocessed data to obtain the P - value and the data sequence corresponding to the P - value; When the P - value of the preprocessed data test ≤ 0.05, take the optimal difference order as 0, and obtain the data sequence at this time as the stable data sequence; when the P - value of the ADF test of the sequence after the first - order difference ≤ 0.05, take the optimal difference order as 1, and obtain the data sequence at this time as the stable data sequence; when the P - value of the ADF test of the sequence after the N - th order difference ≤ 0.05, take the optimal difference order as N, and obtain the data sequence at this time as the stable data sequence; N is a natural number.

5. The prediction method for the cutter head torque parameters of a shield machine according to claim 1, characterized in that The S400 includes: S410. Based on the Bayesian information criterion, traverse the combinations of autoregressive orders and moving average orders within a preset range, and select the combination of autoregressive order and moving average order when the Bayesian information criterion value is the smallest in the stable data sequence; S420. Obtain the optimal ARIMA model according to the difference order corresponding to the stable data sequence and the combination of autoregressive order and moving average order.

6. The prediction method for the cutter head torque parameters of a shield machine according to claim 5, wherein The S410 includes: S411. Within the preset range of autoregressive order ∈ [0, 5] and moving average order ∈ [0, 5], traverse all combinations of autoregressive order and moving average order based on the Bayesian Information Criterion (BIC) to obtain the BIC value for each combination. The BIC value is obtained through the following formula: ； where BIC is the Bayesian Information Criterion value, L is the model likelihood function, representing the goodness of fit for the stable data sequence, n is the number of samples, p is the autoregressive order, and q is the moving average order. S412. Select the combination of autoregressive order and moving average order with the smallest BIC value as the optimal order combination.

7. The prediction method for the cutter head torque parameters of a shield machine according to claim 6, characterized in that The specific steps of S410 include: S413. Divide the preset range of autoregressive order and moving average order into k subtasks, where k is the number of computing nodes. S414. Use the CUDA framework based on GPU to assign each subtask to an independent computing node, and each node calculates the BIC value in parallel based on the assigned order combination range. S415. Use a message queue to collect the BIC values of all computing nodes in real time, and extract the combination of autoregressive order and moving average order corresponding to the minimum BIC value through a reduction operation.

8. The prediction method for the cutter head torque parameters of a shield machine according to claim 3, characterized in that, The method further includes: In a formation with uneven hardness, if the optimal differential order ≥ 2, synchronously increase the sliding window length to 1.5 times the original window.

9. The prediction method for the cutter head torque parameters of a shield machine according to claim 1, wherein, The specific steps of S600 include: S610. When the predicted value of the cutter head torque parameter exceeds 20% of the theoretical torque value under the current geological conditions, reduce the cutter head rotation speed by 0.5 - 1.0 r / min and reduce the tunneling speed to 80% - 90% of the original set value. S620. When the predicted value of the cutter head torque parameter is within the range of ±15% of the theoretical torque value under the current geological conditions, increase the tunneling speed to 105% - 110% of the original set value.

10. The prediction method for the cutter head torque parameters of a shield machine according to claim 3, characterized in that The specific steps of S230 include: S231. Obtain the mean μ and standard deviation σ of the data after removing the data in the empty push section; set the outlier judgment threshold as the interval range from μ - 3σ to μ + 3σ. S232. Traverse the data after removing the data in the empty push section. If any parameter value of the cutter head torque, tunneling speed, or soil chamber pressure exceeds the corresponding threshold interval, it is determined as an outlier and removed to obtain the preprocessed data.

Images