Shield cutter head mud cake prejudgment method based on cutter multi-source monitoring information

Abstract

The invention discloses a shield cutterhead mud cake prejudgment method based on cutter multi-source monitoring information, and belongs to the technical field of shield tunneling machine cutterhead detection.The shield cutterhead mud cake prejudgment method comprises the steps that temperature and vibration frequency of each cutter on a shield cutterhead are collected in real time through a temperature sensor and a vibration sensor, shield tunneling parameters are recorded synchronously, and a multi-modal data set is formed; on the basis of an improved inverse distance weighted interpolation algorithm, predicting temperature-vibration distribution of an area without sensors, and generating a dynamic thermodynamic diagram of the cutterhead; the temperature change rate and the vibration frequency change rate are calculated through an isolated forest algorithm, and when the two-parameter change rate exceeds a preset threshold value, mud cakes are formed, and the machine is stopped for checking; otherwise, the shield tunneling machine continues to conduct tunneling, the preset threshold value is dynamically corrected on the basis of the Bayesian algorithm in combination with newly-collected data, and meanwhile the preset threshold value is optimized through actual inspection result feedback. According to the method, misjudgment caused by single data monitoring and fixed threshold judgment is avoided, and the accuracy of judging mud cakes formed on the shield cutter head is improved.

Description

Technical Field

[0001] The present application relates to the technical field of shield machine cutterhead detection, and in particular to a shield machine cutterhead mud cake prediction method based on multi-source tool monitoring information. Background Art

[0002] In recent years, my country's urbanization has accelerated, and the demand for transportation infrastructure and municipal public service facilities has shown a significant upward trend. Shield construction technology, due to its high safety, strong stability, energy conservation and environmental protection, and outstanding construction efficiency, has become increasingly widely used in underground engineering construction. However, when shield tunneling through clay-rich strata such as strongly or moderately weathered mudstone or argillaceous sandstone, the fine clay particles removed tend to adhere to the cutterhead. Under the influence of earth pressure and frictional heat generated by the cutterhead, a hardened mud cake forms. The formation of this mud cake reduces cutter penetration, thereby reducing shield construction efficiency. It can also cause the cutterhead to become stuck and wear unevenly, damaging the cutterhead structure and seriously threatening project safety.

[0003] During shield construction, as the amount of mud cake on the cutterhead increases, the shield torque and thrust will increase significantly, and the shield advancement speed will slow down. The shield driver can judge that the cutter is in an abnormal state based on the slow advancement speed of the shield. However, this detection method has obvious hysteresis problems. It is often discovered only when the mud cake has accumulated seriously, and it is impossible to accurately locate the specific location where the mud cake is formed. On the other hand, the cutterhead temperature will rise rapidly after the mud cake is formed. The cutterhead mud cake situation can be predicted by monitoring the cutterhead temperature information. However, at this stage, it only relies on single temperature data and ignores multi-dimensional information such as formation data and shield tunneling parameters. The detection accuracy is insufficient, and the temperature threshold mostly relies on manual experience and lacks an adaptive optimization mechanism. Summary of the Invention

[0004] In response to the above-mentioned deficiencies in the prior art, the present application provides a shield cutterhead mud cake prediction method based on multi-source monitoring information of the tool. By monitoring the tool temperature and vibration information, coordinating the excavation parameters to form multimodal data, and combining lightweight algorithms and adaptive optimization mechanisms, efficient and accurate shield cutterhead mud cake judgment can be achieved.

[0005] In order to achieve the above-mentioned invention objectives, the technical solutions adopted in this application are:

[0006] This application provides a shield cutterhead mud cake prediction method based on tool multi-source monitoring information, including:

[0007] S1: Using temperature sensors and vibration sensors to collect the temperature and vibration frequency of each cutter on the shield cutter head in real time, and synchronously record the shield tunneling parameters of the shield cutter head to form a multimodal data set;

[0008] S2: Based on the improved inverse distance weighted interpolation algorithm, the temperature and vibration of each point where the sensor is not installed are predicted according to the temperature and vibration frequency to generate a dynamic thermal map of the cutterhead;

[0009] S3: The temperature change rate and vibration frequency change rate of each point are calculated using the isolation forest algorithm. When the temperature change rate and vibration frequency change rate exceed the preset threshold, the cutter is considered to have a mud cake phenomenon, and the machine is stopped for inspection and enters S4. Otherwise, the shield machine continues to excavate, and the preset threshold is dynamically adjusted using the Bayesian algorithm, and enters S1.

[0010] S4: Based on whether the shutdown inspection result is consistent with the threshold prediction result, if so, the mud cake is cleaned, the shield cutter head continues shield tunneling, and the preset threshold is dynamically adjusted using the Bayesian algorithm, and the process goes to S1; otherwise, the preset threshold is optimized and the process goes to S1.

[0011] Furthermore, the S2 specifically includes:

[0012] S201: Calculating the temperature and vibration frequency of the cutterhead at all positions using the improved inverse distance weighted interpolation algorithm based on the tool temperature and vibration frequency of the shield cutterhead during operation collected in S1;

[0013] S202: Perform data normalization on the calculated temperatures and vibration frequencies of the cutter head at all positions. The data normalization formula is:

[0014]

[0015] Among them, u is the mean, σ is the standard deviation, X is the original sample data, X norm To standardize sample data;

[0016] S203: normalizing the standardized sample data, and mapping the normalized temperature-vibration data through the HSL color space;

[0017] S204: Convert the HSL value of each data point into RGB color, map the area of the cutter head to the corresponding color according to the coordinates, and generate a dynamic heat map of the cutter head.

[0018] Furthermore, the improved inverse distance weighted interpolation algorithm:

[0019]

[0020] Among them, T i is the temperature, V i is the vibration frequency, d i is the distance, p is the attenuation coefficient, x and y represent the positions of the cutter head, i is the monitoring point, and n is the total monitoring points.

[0021] Furthermore, the S3 specifically includes:

[0022] S301: extracting subsamples from the normalized data and constructing an isolated tree;

[0023] S302: Form an isolation forest based on the constructed isolation trees; for each subsample, traverse each tree and record the path length from the root node to the leaf node;

[0024] S303: Calculate the average path length in all trees and normalize it to anomaly score s:

[0025]

[0026] Where a is the sample point, m is the total number of samples, c(.) is the path length correction term, h(.) is the path length from the root node to the leaf node, and E(.) is the average path length;

[0027] S304: Selecting a data position with an abnormality score s>0.5, and calculating the temperature change rate and vibration frequency change rate of the cutter head position corresponding to the data position;

[0028] S305: Compare the temperature change rate and vibration frequency change rate with the preset threshold value. When the temperature change rate and vibration frequency change rate of a certain tool are greater than the temperature threshold value and the vibration threshold value, it is determined that the tool has a mud cake phenomenon, and the machine is stopped for inspection and enters S4; otherwise, the shield machine continues to excavate, and the preset threshold value is dynamically adjusted using the Bayesian algorithm, and enters S1.

[0029] Furthermore, the dynamically adjusting the preset threshold using the Bayesian algorithm specifically includes:

[0030] A1: Obtain historical data and real-time data collected from the shield cutterhead;

[0031] A2: Construct a Bayesian linear regression model based on the data:

[0032] ΔT=β0+β1·X 地层类型 +β2·X 掘进参数 +β3·X 历史误报率 +ε T

[0033] ΔV=γ0+γ1·X 地层类型 +γ2·X 掘进参数 +γ3·X 历史误报率 +ε V

[0034] Among them, ΔT and ΔV are the temperature threshold adjustment values and vibration threshold adjustment values, β0 and γ0 are the initial parameter weight coefficients, β1 and γ1 are the geological type weight coefficients, β2 and γ2 are the excavation parameter weight coefficients, β3 and γ3 are the historical false alarm rate weight coefficients, and ε T , ε V is the noise term, X 地质类型 、X 掘进参数 、X 历史误报率 is the type parameter during shield tunneling;

[0035] A3: Using Markov Chain Monte Carlo to estimate the posterior distribution to determine the weight coefficient;

[0036] A4: Obtaining a current temperature threshold adjustment value and a vibration threshold adjustment value based on the weight coefficient, the tunneling parameter, and the Bayesian linear regression model;

[0037] A5: Generate a current temperature threshold and a vibration threshold according to the basic threshold and the threshold adjustment value.

[0038] Furthermore, the optimizing the preset threshold includes:

[0039] When the shutdown inspection result does not match the threshold judgment result, modify the historical false alarm rate weight coefficient.

[0040] The beneficial effects of this application are: by monitoring multimodal data such as temperature and vibration to predict the occurrence of cutterhead mud cake, the deviation of single data is reduced, the detection reliability is improved, and lightweight algorithms such as isolation forest are used to reduce the relevant calculations of temperature and vibration data, thereby improving computational efficiency. At the same time, the Bayesian algorithm is used to dynamically adjust the preset threshold, reducing manual intervention, thereby improving the accuracy of shield cutterhead mud cake detection. In addition, if the system makes a misjudgment or omission, it can adaptively adjust the preset threshold, thereby improving the accuracy of cutterhead mud cake detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other embodiments can also be obtained based on these drawings.

[0042] Figure 1 A flow chart of a method for predicting mud cake on a shield cutterhead based on multi-source tool monitoring information provided in an embodiment of the present application.

[0043] Figure 2 A schematic diagram of a cutter disc information monitoring system provided in an embodiment of the present application.

[0044] Figure 3 A dynamic thermal diagram of a cutter disc is provided in an embodiment of the present application.

[0045] Among them: 1-shield machine, 2-cutterhead, 3-tool, 4-temperature sensor, 5-vibration sensor, 6-wireless signal receiving module, 7-shield main control room, 8-shield host control system. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field based on this application are within the scope of protection of this application.

[0047] The temperature-based method for monitoring cutterhead mud cake only relies on single temperature data, ignoring multi-dimensional information such as formation data and shield tunneling parameters. The detection accuracy is insufficient. In order to improve the accuracy of cutterhead mud cake judgment, cutterhead mud cake judgment can be performed based on the temperature and vibration data of the tool.

[0048] It is understandable that when mud cakes form, friction increases and the temperature rises in a step-by-step manner. The accumulation of mud cakes causes uneven force on the cutterhead, and the vibration spectrum energy is concentrated in the low-frequency band. There is a time lag between temperature increase and vibration anomalies. Usually, vibration anomalies occur earlier than temperature changes. Through the joint anomaly detection of temperature and vibration signals, the early characteristics of mud cakes on the cutterhead can be identified, the monitoring accuracy can be improved, the normal operation of the shield cutterhead can be ensured, and the operation and maintenance costs can be reduced.

[0049] Based on this, the embodiment of the present application provides a shield cutterhead mud cake prediction method based on tool multi-source monitoring information. This method can be found in Figure 1 , Figure 1 The figure shows a flow chart of a method for predicting mud cake on a shield cutterhead based on multi-source tool monitoring information provided by an embodiment of the present application, including:

[0050] S1: Use temperature sensors and vibration sensors to collect the temperature and vibration frequency of each cutter on the shield cutter head in real time, and simultaneously record the shield tunneling parameters of the shield cutter head to form a multimodal data set.

[0051] In one possible embodiment, the cutterhead information monitoring system can refer to Figure 2 , Figure 2A schematic diagram of a cutterhead information monitoring system provided in an embodiment of the present application is specifically as follows: a temperature sensor 4 and a vibration sensor 5 are installed on the inner wall of each tool 3 of the cutterhead 2, and the temperature sensor 4 and the vibration sensor 5 have built-in wireless transmission modules, so that real-time monitored temperature and vibration data can be wirelessly transmitted; a wireless signal receiving module 6 is installed on the inner side of the shield cutterhead 2 for receiving temperature and vibration data; the wireless signal receiving module 6 transmits the acquired monitoring data to the host control system 8 of the shield main control room 7 through wired transmission.

[0052] In one possible embodiment, shield tunneling parameters at the time of sensor acquisition, including shield thrust, torque, and tunneling speed, are synchronously recorded to form a multimodal data set.

[0053] S2: Based on the improved inverse distance weighted interpolation algorithm, the temperature and vibration of each point where the sensor is not installed are predicted according to the temperature and vibration frequency to generate a dynamic thermal map of the cutterhead.

[0054] Furthermore, the S2 specifically includes:

[0055] S201: Calculating the temperature and vibration frequency of the cutterhead at all positions using the improved inverse distance weighted interpolation algorithm based on the tool temperature and vibration frequency of the shield cutterhead during operation collected in S1;

[0056] S202: Perform data normalization on the calculated temperatures and vibration frequencies of the cutter head at all positions. The data normalization formula is:

[0057]

[0058] Among them, u is the mean, σ is the standard deviation, X is the original sample data, X norm To standardize sample data;

[0059] S203: normalizing the standardized sample data, and mapping the normalized temperature-vibration data through the HSL color space;

[0060] S204: Convert the HSL value of each data point into RGB color, map the area of the cutter head to the corresponding color according to the coordinates, and generate a dynamic heat map of the cutter head. Figure 3 shown.

[0061] Among them, the improved inverse distance weighted interpolation algorithm in S2 is:

[0062]

[0063] Among them, T i is the temperature, V i is the vibration frequency, d iis the distance, p is the attenuation coefficient, x and y represent the positions of the cutter head, i is the monitoring point, and n is the total monitoring points.

[0064] In a possible embodiment, an RGB color system is used to adjust the R, G, and B values according to the temperature. For example, if an RGB color scheme is used, the higher the temperature, the redder the part, and the lower the temperature, the greener the part. The temperature gradually changes from high to low, from red to yellow, and then to green. Under visual observation, it can be judged whether the tool is abnormal based on the dynamic thermal map of the tool disc.

[0065] S3: The temperature change rate and vibration frequency change rate of each point are calculated using the isolation forest algorithm. When the temperature change rate and vibration frequency change rate exceed the preset threshold, the cutter is considered to have a mud cake phenomenon, and the machine is stopped for inspection and enters S4. Otherwise, the shield machine continues to excavate, and the preset threshold is dynamically adjusted using the Bayesian algorithm, and enters S1.

[0066] Furthermore, the S3 specifically includes:

[0067] S301: extracting subsamples from the normalized data and constructing an isolated tree;

[0068] S302: Form an isolation forest based on the constructed isolation trees; for each subsample, traverse each tree and record the path length from the root node to the leaf node;

[0069] S303: Calculate the average path length in all trees and normalize it to anomaly score s:

[0070]

[0071] Where a is the sample point, m is the total number of samples, c(.) is the path length correction term, h(.) is the path length from the root node to the leaf node, and E(.) is the average path length;

[0072] S304: Selecting a data position with an abnormality score s>0.5, and calculating the temperature change rate and vibration frequency change rate of the cutter head position corresponding to the data position;

[0073] S305: Compare the temperature change rate and vibration frequency change rate with the preset threshold value. When the temperature change rate and vibration frequency change rate of a certain tool are greater than the temperature threshold value and the vibration threshold value, it is determined that the tool has a mud cake phenomenon, and the machine is stopped for inspection and enters S4; otherwise, the shield machine continues to excavate, and the preset threshold value is dynamically adjusted using the Bayesian algorithm, and enters S1.

[0074] In one possible embodiment, based on the tool temperature and vibration frequency of the shield cutterhead during operation collected in step S1, the IDW interpolation method is used to calculate the temperature and vibration frequency of the cutterhead at all positions; (2) the temperature and vibration frequency are standardized and normalized; (3) subsamples are extracted from the normalized data to construct an isolation tree, and each isolation tree recursively partitions the data until a termination condition is met. The termination condition is generally: only one sample is left at the node, or the height of the tree reaches a preset limit, usually log2(α), where α is the total number of subsamples; (4) Construct 100 isolated trees to form an isolation forest; for each data point α, traverse each tree and record the path length h(α) from the root node to the leaf node; (5) Calculate the average path length E(h(α)) in all trees and standardize it to the anomaly score s, s(α,n)=2-E(h(α)) / c(n), c(n) is the path length correction term; s≈1: highly likely anomaly point, s≤0.5, normal sample; (6) Select the data position where s>0.5 and calculate the temperature change rate and vibration frequency change rate at that position, ΔT=(T t -T0) / t, ΔV=(V t -V0) / t, where T t is the temperature during the time period t, T0 is the initial temperature, V t is the vibration frequency in the time period t, and V0 is the initial vibration frequency; (7) ΔT>ΔT0, ΔV>ΔV0 is the position where the cutterhead mud cake is determined, where ΔT0 and ΔV0 are the temperature threshold and vibration threshold, respectively.

[0075] Furthermore, the dynamically adjusting the preset threshold using the Bayesian algorithm specifically includes:

[0076] A1: Obtain historical data and real-time data collected from the shield cutterhead;

[0077] A2: Construct a Bayesian linear regression model based on the data:

[0078] ΔT=β0+β1·X 地层类型 +β2·X 掘进参数 +β3·X 历史误报率 +ε T

[0079] ΔV=γ0+γ1·X 地层类型 +γ2·X 掘进参数 +γ3·X 历史误报率 +ε V

[0080] Among them, ΔT and ΔV are the temperature threshold adjustment values and vibration threshold adjustment values, β0 and γ0 are the initial parameter weight coefficients, β1 and γ1 are the geological type weight coefficients, β2 and γ2 are the excavation parameter weight coefficients, β3 and γ3 are the historical false alarm rate weight coefficients, and ε T , ε V is the noise term, X 地质类型 、X 掘进参数 、X 历史误报率 is the excavation parameter during shield tunneling;

[0081] A3: Using Markov Chain Monte Carlo to estimate the posterior distribution to determine the weight coefficient;

[0082] A4: Obtaining a current temperature threshold adjustment value and a vibration threshold adjustment value based on the weight coefficient, the tunneling parameter, and the Bayesian linear regression model;

[0083] A5: Generate a current temperature threshold and a vibration threshold according to the basic threshold and the threshold adjustment value.

[0084] In one possible embodiment, (1) the formation type data is one-hot encoded, and the excavation parameters and historical false alarm rates are annotated; (2) a Bayesian linear regression model is used to obtain the temperature threshold adjustment value ΔT and the vibration threshold adjustment value ΔV:

[0085] ΔT=β0+β1·X 地层类型 +β2·X 掘进参数 +β3·X 历史误报率 +ε T

[0086] ΔV=γ0+γ1·X 地层类型 +γ2·X 掘进参数 +γ3·X 历史误报率 +ε V

[0087] Set the weight coefficient β i , γ i Conforms to Gaussian distribution;

[0088] (3) The posterior distribution is estimated using Markov chain Monte Carlo. (4) The mean of each weight coefficient is extracted from the posterior distribution, combined with the excavation parameters in step S1, and substituted into the regression model to obtain the current temperature threshold and vibration threshold adjustment values, which are then superimposed on the basic threshold to generate the current temperature threshold and vibration threshold as ΔT0+ΔT and ΔV0+ΔV, respectively.

[0089] S4: Based on whether the shutdown inspection result is consistent with the threshold prediction result, if so, the mud cake is cleaned, the shield cutter head continues shield tunneling, and the preset threshold is dynamically adjusted using the Bayesian algorithm, and the process goes to S1; otherwise, the preset threshold is optimized and the process goes to S1.

[0090] Furthermore, the optimizing the preset threshold includes:

[0091] When the shutdown inspection result does not match the threshold judgment result, modify the historical false alarm rate weight coefficient.

[0092] In one possible embodiment, if the shutdown inspection result is consistent with the threshold prediction result, it means that the prediction is correct and the tool has indeed formed a mud cake. In this case, the mud cake is cleaned, the shield cutter head continues shield tunneling, and the shield tunneling parameters of the shield cutter head are recorded to form new multimodal data. Based on the new multimodal data, the preset threshold is dynamically adjusted using the Bayesian algorithm; otherwise, it means that the prediction is inaccurate and the preset threshold is optimized.

[0093] In one possible embodiment, when the system detects a false positive, that is, the threshold judgment result determines that a mud cake phenomenon has occurred, but the shutdown inspection result does not determine that a mud cake has occurred, it is a false positive, and the historical false alarm rate weight coefficients β3 and γ3 are reduced, thereby reducing the threshold sensitivity; when the system detects a missed alarm, that is, the threshold judgment result does not determine that a mud cake phenomenon has occurred, but the shutdown inspection result determines that a mud cake has occurred, then the historical false alarm rate weight coefficient is increased, thereby improving the threshold sensitivity.

[0094] This application predicts cutterhead mud cake by monitoring multimodal data such as temperature and vibration, reducing single-data bias and improving detection reliability. It also utilizes lightweight algorithms such as isolation forests to reduce the associated calculations for temperature and vibration data, improving computational efficiency. Furthermore, a Bayesian algorithm is used to dynamically adjust preset thresholds, reducing manual intervention and thus improving the accuracy of shield cutterhead mud cake detection. Furthermore, if the system misjudges or misses a detection, it can adaptively adjust the preset thresholds, improving the accuracy of cutterhead mud cake detection.

[0095] Each embodiment in this specification is described in a related manner. Similar parts between the embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences from other embodiments. The above description is only a preferred embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of this application are included in the scope of protection of this application.

Claims

1. A shield cutterhead mud cake prediction method based on tool multi-source monitoring information, characterized in that: include: S1: Using temperature sensors and vibration sensors to collect the temperature and vibration frequency of each cutter on the shield cutter head in real time, and synchronously record the shield tunneling parameters of the shield cutter head to form a multimodal data set; S2: Based on the improved inverse distance weighted interpolation algorithm, the temperature and vibration of each point where the sensor is not installed are predicted according to the temperature and vibration frequency to generate a dynamic thermal map of the cutterhead; S3: The temperature change rate and vibration frequency change rate of each point are calculated using the isolation forest algorithm. When the temperature change rate and vibration frequency change rate exceed the preset threshold, the cutter is considered to have a mud cake phenomenon, and the machine is stopped for inspection and enters S4. Otherwise, the shield machine continues to excavate, and the preset threshold is dynamically adjusted using the Bayesian algorithm, and enters S1. S4: Based on whether the shutdown inspection result is consistent with the threshold prediction result, if so, the mud cake is cleaned, the shield cutter head continues shield tunneling, and the preset threshold is dynamically adjusted using the Bayesian algorithm, and the process goes to S1; otherwise, the preset threshold is optimized and the process goes to S1.

2. The shield cutterhead mud cake prediction method based on tool multi-source monitoring information according to claim 1 is characterized in that: The S2 specifically includes: S201: Calculating the temperature and vibration frequency of the cutterhead at all positions using the improved inverse distance weighted interpolation algorithm based on the tool temperature and vibration frequency of the shield cutterhead during operation collected in S1; S202: Perform data normalization on the calculated temperatures and vibration frequencies of the cutter head at all positions. The data normalization formula is: Among them, u is the mean, σ is the standard deviation, X is the original sample data, X norm To standardize sample data; S203: normalizing the standardized sample data, and mapping the normalized temperature-vibration data through the HSL color space; S204: Convert the HSL value of each data point into RGB color, map the area of the cutter head to the corresponding color according to the coordinates, and generate a dynamic heat map of the cutter head.

3. The shield cutterhead mud cake prediction method based on tool multi-source monitoring information according to claim 2 is characterized in that: The improved inverse distance weighted interpolation algorithm: Among them, T i is the temperature, V i is the vibration frequency, d i is the distance, p is the attenuation coefficient, x and y represent the positions of the cutter head, i is the monitoring point, and n is the total monitoring points.

4. The shield cutterhead mud cake prediction method based on tool multi-source monitoring information according to claim 1 is characterized in that: The S3 specifically includes: S301: extracting subsamples from the normalized data and constructing an isolated tree; S302: Form an isolation forest based on the constructed isolation trees; for each subsample, traverse each tree and record the path length from the root node to the leaf node; S303: Calculate the average path length in all trees and normalize it to anomaly score s: Where a is the sample point, m is the total number of samples, c(.) is the path length correction term, h(.) is the path length from the root node to the leaf node, and E(.) is the average path length; S304: Selecting a data position with an abnormality score s>0.5, and calculating the temperature change rate and vibration frequency change rate of the cutter head position corresponding to the data position; S305: Compare the temperature change rate and vibration frequency change rate with the preset threshold value. When the temperature change rate and vibration frequency change rate of a certain tool are greater than the temperature threshold value and the vibration threshold value, it is determined that the tool has a mud cake phenomenon, and the machine is stopped for inspection and enters S4; otherwise, the shield machine continues to excavate, and the preset threshold value is dynamically adjusted using the Bayesian algorithm, and enters S1.

5. The shield cutterhead mud cake prediction method based on tool multi-source monitoring information according to claim 1 or 4 is characterized in that: The dynamically adjusting the preset threshold using the Bayesian algorithm specifically includes: A1: Obtain historical data and real-time data collected from the shield cutterhead; A2: Construct a Bayesian linear regression model based on the data: ΔT=β0+β1·X 地层类型 +β2·X 掘进参数 +β3·X 历史误报率 +e T ΔV=γ0+γ1·X 地层类型 +γ2·X 掘进参数 +γ3·X 历史误报率 +e V Among them, ΔT and ΔV are the temperature threshold adjustment values and vibration threshold adjustment values, β0 and γ0 are the initial parameter weight coefficients, β1 and γ1 are the geological type weight coefficients, β2 and γ2 are the excavation parameter weight coefficients, β3 and γ3 are the historical false alarm rate weight coefficients, and ε T , ε V is the noise term, X 地质类型 、X 掘进参数 、X 历史误报率 is the type parameter during shield tunneling; A3: Using Markov Chain Monte Carlo to estimate the posterior distribution to determine the weight coefficient; A4: Obtaining a current temperature threshold adjustment value and a vibration threshold adjustment value based on the weight coefficient, the tunneling parameter, and the Bayesian linear regression model; A5: Generate a current temperature threshold and a vibration threshold according to the basic threshold and the threshold adjustment value.

6. The shield cutterhead mud cake prediction method based on tool multi-source monitoring information according to claim 1 is characterized in that: The optimizing the preset threshold comprises: When the shutdown inspection result does not match the threshold judgment result, modify the historical false alarm rate weight coefficient.

Images