Shield tunneling machine cutter state analysis and diagnosis method and system

Abstract

The invention relates to the technical field of cutter state analysis, in particular to a shield tunneling machine cutter state analysis and diagnosis method and system.The shield tunneling machine cutter state analysis and diagnosis system comprises a signal obtaining module, a processing module and an image processing module, the signal obtaining module comprises two vibration sensors and a three-phase Hall power sensor, and the processing module comprises an industrial personal computer; the image processing module comprises a high-definition camera, the high-definition camera is in signal connection with an industrial personal computer, and the industrial personal computer is used for conducting sharpening processing on shield tunneling machine cutter images shot by the high-definition camera. According to the shield tunneling machine cutter state analysis and diagnosis method and system, the vibration signal and the power signal which can fully reflect the cutter wear state are selected for signal detection, the vibration signal is decomposed and reconstructed, the cutter wear characteristic signal is extracted, the vibration characteristic signal and the power characteristic signal are fused by adopting a fuzzy data fusion method, and the cutter wear state is analyzed and diagnosed. The limitation that a single characteristic signal provides the cutter state information is avoided, and the more comprehensive and accurate cutter state condition is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of tool state analysis, and in particular to a shield machine tool state analysis and diagnosis method and system. Background Art

[0002] For a shield machine, the cutter is one of the most important parts of its work and an important tool for the advancement of the shield machine. When the cutter is damaged or in a bad state, the shield machine needs to stop and change the cutter, which will cause huge economic losses and even safety issues. The collapse of the cutter may even damage the cutter barrel or even the cutterhead of the shield machine, causing irreparable losses. In addition, the cutter wear detection signal is a nonlinear signal, and the nonlinear state signal is often a complex signal that is non-Gaussian, non-stationary, and strongly coupled. This makes it difficult to accurately extract and identify the state features, especially under strong interference and multiple working conditions. The problem is particularly prominent. On the shield machine, the cutter is placed in the mud and the working environment is harsh, which places extremely high demands on the reliability of the detection system. The accuracy of the cutter state diagnosis depends largely on the effectiveness, completeness, and accuracy of the state feature extraction of the target system.

[0003] At present, the status of shield machine cutters mainly focuses on cutter wear, cutter tooth collapse, cutter bolt collapse and other problems. In response to such problems, domestic and foreign researchers have used the following methods: The research on cutter wear detection by shield industry practitioners at home and abroad has never stopped. Shield cutter wear detection is divided into three types according to the detection method: electrical detection, hydraulic detection, and gas detection. According to the detection characteristics, it is divided into two types: limit detection and continuous detection. Limit detection means that when the cutter wears to a certain value, an alarm signal is generated. At present, electrical detection, hydraulic detection, and gas detection can all be used as limit detection. For continuous detection, only electrical detection can be used as continuous detection. The above wear detection technologies cannot take into account both multi-point detection and accuracy and reliability.

[0004] How to pre-process and deeply mine data to completely and accurately extract status features is an urgent problem that needs to be solved. How to avoid the limitations of a single characteristic signal in providing tool status information to obtain a more comprehensive and accurate tool cutting status is also a difficulty that needs to be solved urgently in this field. Summary of the invention

[0005] The purpose of the present invention is to provide a shield machine cutter state analysis and diagnosis method and system to solve the problems raised in the above background technology.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a shield machine tool state analysis and diagnosis system, comprising a signal acquisition module, a processing module and an image processing module, the signal acquisition module comprising two vibration sensors and a three-phase Hall power sensor, the processing module comprising an industrial computer, the industrial computer is connected to the vibration sensor and the three-phase Hall power sensor signals, and the vibration signal is subjected to multi-resolution analysis by the Mallat fast algorithm, and is decomposed and single-branch reconstructed at 5 scales to obtain low-frequency signals and high-frequency signals at 5 scales as characteristic signals of tool wear.

[0007] The image processing module includes a high-definition camera, which is connected to an industrial computer by signal. The industrial computer is used to perform clarity processing on the shield machine tool image taken by the high-definition camera.

[0008] Preferably, the industrial computer matches the characteristic signal of tool wear with the shield machine tool image, and analyzes the shield machine tool status.

[0009] Preferably, a diagnostic method for a shield machine tool state analysis and diagnosis system comprises the following steps:

[0010] S1: The vibration sensor and the three-phase Hall power sensor obtain the vibration data, speed data and input power signal of the shield machine tool and the main shaft drive motor. After the signal is transmitted to the industrial computer, the vibration signal is analyzed by the Mallat fast algorithm at multiple resolutions, and decomposed and reconstructed at five scales to obtain low-frequency and high-frequency signals at five scales as the characteristic signals of tool wear.

[0011] S2: Two vibration sensors and three-phase Hall power sensors obtain a total of 9 groups of data. Fuzzy set theory uses membership functions to represent the uncertainty of each sensor information. Then, fuzzy transformation is used for data processing. Multi-sensor fuzzy fusion theory is applied to fuse the three groups of membership values. After the three sensors are fused, the membership values ​​of the three states of the tool being measured are obtained. Then, the state of the tool being measured is determined according to the decision rule.

[0012] S3: The shield machine tool image taken by the high-definition camera is processed for clarity, the characteristic signal of tool wear by the industrial computer is matched with the shield machine tool image, and the shield machine tool status is manually analyzed.

[0013] Preferably, the signal acquisition parameters in S1 are: the acquisition frequencies of the vibration signal and the power signal are both 10 kHz.

[0014] Preferably, the fusing process of the membership values ​​in S2 comprises the following steps:

[0015] S2.1: Assume that the tool state is given by uj (j=1,2,3) indicates that the sensor is represented by v i (i=1,2,3) indicates that the sensor v i Inferred to be state u j The probability is r i j, that is, v i To u j The membership degree can be constructed by constructing the fuzzy relationship matrix R = (r i j) m×n ;

[0016] S2.2: The contribution of each sensor to the tool state decision is called sensor weight, expressed as a i (i=1,2,3) indicates that the fuzzy vector A=(a1,a2,a3) indicates the credibility of the judgment of various sensors;

[0017] S2.3: The possibility of the tool state obtained by fuzzy transformation B = A·R = (b1, b2, b3), b j (j=1,2,3) is the membership degree R of the fuzzy subset obtained by comprehensive judgment of the tool state.

[0018] Preferably, the fuzzy relationship matrix R must determine the form of the membership function, that is, determine the membership function form of the characteristic values ​​observed by each sensor relative to each state of tool wear, and divide the tool wear state into three fuzzy states (initial, normal, and acute).

[0019] Preferably, the decision rule also needs to exclude the probability of false alarm rate and missed alarm rate detected by the sensor when determining.

[0020] Preferably, the method for clearing the image in S3 is to perform random amplitude enhancement and attenuation after slicing.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] 1. The shield machine tool state analysis and diagnosis method and system selects vibration signals and power signals that can fully reflect the tool wear state for signal detection, uses orthogonal wavelet transform technology to decompose and reconstruct the vibration signal, extracts the tool wear characteristic signal, and uses fuzzy data fusion method to fuse the vibration and power characteristic signals, avoiding the limitation of a single characteristic signal in providing tool state information and obtaining a more comprehensive and accurate tool state.

[0023] 2. The shield machine tool state analysis and diagnosis method and system, the industrial computer matches the characteristic signal of tool wear with the shield machine tool image, and analyzes the shield machine tool state, avoiding the limitation of a single characteristic signal providing tool state information. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0025] Figure 1 This is a flow chart of the shield machine cutter state characteristic signal fusion of the present invention. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0028] See also Figure 1 , the present invention provides a technical solution:

[0029] A shield machine tool state analysis and diagnosis system comprises a signal acquisition module, a processing module and an image processing module. The signal acquisition module comprises two vibration sensors and a three-phase Hall power sensor. The processing module comprises an industrial computer. The industrial computer is connected to the vibration sensor and the three-phase Hall power sensor signals. The vibration signal is subjected to multi-resolution analysis by a Mallat fast algorithm, and is decomposed and reconstructed at five scales to obtain low-frequency signals and high-frequency signals at five scales as characteristic signals of tool wear.

[0030] The image processing module includes a high-definition camera, which is connected to an industrial computer by signal. The industrial computer is used to perform clarity processing on the shield machine tool image taken by the high-definition camera.

[0031] The industrial computer matches the characteristic signal of tool wear with the shield machine tool image and analyzes the status of the shield machine tool.

[0032] A diagnostic method for a shield machine tool state analysis and diagnosis system comprises the following steps:

[0033] S1: The vibration sensor and the three-phase Hall power sensor obtain the vibration data, speed data and input power signal of the shield machine tool and the main shaft drive motor. After the signal is transmitted to the industrial computer, the vibration signal is analyzed by the Mallat fast algorithm at multiple resolutions, and decomposed and reconstructed at five scales to obtain low-frequency and high-frequency signals at five scales as the characteristic signals of tool wear.

[0034] S2: Two vibration sensors and three-phase Hall power sensors obtain a total of 9 groups of data. Fuzzy set theory uses membership functions to represent the uncertainty of each sensor information. Then, fuzzy transformation is used for data processing. Multi-sensor fuzzy fusion theory is applied to fuse the three groups of membership values. After the three sensors are fused, the membership values ​​of the three states of the tool being measured are obtained. Then, the state of the tool being measured is determined according to the decision rule.

[0035] S3: The shield machine tool image taken by the high-definition camera is processed for clarity, the characteristic signal of tool wear by the industrial computer is matched with the shield machine tool image, and the shield machine tool status is manually analyzed.

[0036] Vibration sensors and three-phase Hall power sensors are properly installed on the shield machine. The vibration sensor should be installed close to the tool to accurately obtain the tool's vibration data; the three-phase Hall power sensor is installed on the power line of the shield machine's main shaft drive motor to measure the motor's input power signal.

[0037] The original vibration signal of the tool (set as x(t)) and the input power signal of the motor (set as y(t)) are obtained through these sensors, and the speed data of the tool is obtained through a suitable speed measurement device. The speed data can be obtained through an encoder and other devices and recorded synchronously with the vibration and power signals.

[0038] The collected vibration signal x(t) is denoised to remove high-frequency noise and interference signals. Filtering techniques such as Butterworth filter or wavelet threshold denoising can be used.

[0039] The power signal y(t) is subjected to the same preprocessing to ensure the signal quality for subsequent analysis.

[0040] The processed vibration signal and power signal are transmitted to the industrial computer through data acquisition card and other devices.

[0041] Mallat fast algorithm multi-resolution analysis stage (taking vibration signal as an example)

[0042] Wavelet basis function selection:

[0043] According to the characteristics of the vibration signal, the appropriate Daubechies wavelet basis function is selected, and the vibration signal x(t) is decomposed into multi-resolution using the Mallat algorithm on an industrial computer. The Mallat algorithm is a fast decomposition algorithm based on wavelet transform, and its basic principle is to decompose the signal into approximate signals and detail signals of different scales (resolutions).

[0044] Assume that the number of decomposition layers is J. For a discrete signal x(n), where n is a discrete time point), after decomposition using the Mallat algorithm, the approximate coefficient c can be obtained. j,k and detail factor d j,k , where j = 1, 2, …, J represents the decomposition scale and k represents the discrete time point at this scale.

[0045] The specific decomposition process is as follows:

[0046] First, the first layer of decomposition is performed to decompose the signal x(n) into an approximate signal c 1,k (low frequency part) and detail signal d 1,k (high frequency part).

[0047] Then, the approximate signal c 1,k As the input of the next layer, continue to decompose to get c 2,k and d 2,k , and so on, until the predetermined decomposition level J is reached.

[0048] Multi-resolution analysis is performed on the vibration signal based on the approximate coefficients and detail coefficients obtained by decomposition.

[0049] At different decomposition scales, the approximation coefficient c j,k Represents the low-frequency component of the signal at this scale, reflecting the general outline or trend of the signal; detail coefficient d j,k It represents the high-frequency components of the signal at this scale and reflects the detailed changes of the signal, such as mutations and shocks.

[0050] For example, the detail coefficient at a lower scale (smaller j) can capture the high-frequency impact signal in the tool vibration, which may be related to the collision between the tool and the rock; the approximate coefficient at a higher scale (larger j) can reflect the long-term trend of the tool vibration, such as the overall change of vibration during the advancement of the shield machine.

[0051] Feature extraction and analysis:

[0052] Extract valuable feature information from the decomposed coefficients. Energy features at different scales can be calculated, such as approximate signal energy and detail signal energy

[0053] By analyzing the distribution of these energies at different scales, we can understand the frequency distribution characteristics of the tool vibration signal. For example, if the energy of the detail signal suddenly increases at a certain scale, it may indicate that the tool has abnormal vibration within this frequency range, such as changes in vibration frequency caused by wear, breakage, etc.

[0054] At the same time, the speed data and power signal can be combined for comprehensive analysis. For example, when the tool speed changes, the multi-resolution characteristics of the vibration signal change, as well as the relationship between the motor input power and vibration, to determine whether the shield machine is working normally.

[0055] The captured images are transmitted to the image processor, which performs VLAD fusion on the image data to obtain vectors. First, the features of each frame of a video are clustered to obtain multiple cluster centers, and all features are assigned to the specified cluster centers. The feature vectors in each cluster area are averaged, and finally the feature vectors of all cluster areas are merged as the feature vector of the entire video. The feature vectors of all frames are clustered to one center point, and VLAD clusters the feature vectors of all frames to multiple center points. By stacking all feature vectors, richer features can be obtained with less information loss. The obtained vectors are used to model the image through the low-rank approximation model in 3D convolution to obtain feature results.

[0056] A 3×3×3 convolution kernel is used in all images, and the convolution layer stride is set to 1. The input of VGG is set to an RGB image of size 224×244. The RGB mean is calculated for all images in the training set, and then the image is passed as input to the VGG convolutional network.

[0057] Assuming that n recognition points are extracted from an image, the visual features of the image can be expressed as χ = {x1, ..., x j ,...x n}, the corresponding position feature can be expressed as P = P = {P1, ..., P j , ..., P n}, where x j and P j The distribution represents the visual features and position features of the jth recognition point. In order to obtain a more efficient recognition representation, it is necessary to encode the underlying image features to obtain a vector-level representation of the image.

[0058] First, k-means clustering is used to cluster visual features to generate a visual dictionary. Assuming that the size of the visual dictionary VD1 obtained by clustering is k1, VD1 is expressed as VD1 = {C1, ..., C i , …, C k1}, where C i Represents the i-th cluster center in the visual dictionary. The original VLAD encoding is to calculate the difference between each cluster center and its elements, that is, the encoding vector of the i-th cluster center is expressed as:

[0059] where x j Represents the cluster center C i The j-th visual feature contained, N i Represents the cluster center C i The number of visual features included. In order to enhance the role of cluster centers in visual features, the calculation of the residual between the cluster center and its most similar element is added. The VLAD calculation formula of the i-th cluster center is:

[0060] where x t Represents the cluster center C i The most similar visual feature among the contained visual features. In summary, when the size of the visual dictionary is k1, the image encoding vector of the image can be expressed as:

[0061]

[0062] The image data is enhanced and the amplitude is randomly enhanced. After the effective image data is extracted, the data x is obtained. valid , the number of sampling points is N valid , with N clip (e.g. 22050) is the number of sampling points for each image, for x valid Slice to get multiple slice data x clip (λ c ), where λ c The slice number.

[0063] For each slice data x clip (λ c ) performs random amplitude enhancement or attenuation. First, a gain value rn_s is randomly selected from [rn_l, rn_h] (for example, [-10, 10]), and then a second random value is selected from [rn_s-rn_o, rn_h+rn_o] as the gain of a slice, and directly multiplied with the slice data to obtain the slice data after gain processing. The gain values ​​of all slices are randomly obtained in the range of [rn_s-rn_o, rn_h+rn_o] and gain processed in the same way. Among them, rn_l and rn_h are the lower and upper limits of the first-level random gain, respectively, and rn_o is the offset of the second-level gain. The data slices after random gain processing are resynthesized into complete image data.

[0064] The signal acquisition parameters in S1 are: the acquisition frequency of vibration signal and power signal are both 10kHz.

[0065] The fusion processing of the membership values ​​in S2 includes the following steps:

[0066] S2.1: Assume that the tool state is given by u j (j=1,2,3) indicates that the sensor is represented by v i (i=1,2,3) indicates that the sensor v i Inferred to be state u j The probability is r i j, that is, v i To u j The membership degree can be constructed by constructing the fuzzy relationship matrix R = (r i j) m×n ;

[0067] S2.2: The contribution of each sensor to the tool state decision is called sensor weight, expressed as a i (i=1,2,3) indicates that the fuzzy vector A=(a1,a2,a3) indicates the credibility of the judgment of various sensors;

[0068] S2.3: The possibility of the tool state obtained by fuzzy transformation B = A·R = (b1, b2, b3), b j (j=1,2,3) is the membership degree R of the fuzzy subset obtained by comprehensive judgment of the tool state.

[0069] The fuzzy relationship matrix R must determine the form of the membership function, that is, determine the membership function form of the eigenvalues ​​observed by each sensor relative to each state of tool wear, and divide the tool wear state into three fuzzy states (initial, normal, and acute).

[0070] H1, H2, and H3 are used to represent them respectively, and i and j are used to represent the wear states of different sensors and tools respectively. According to the observation of characteristic signal data, the membership function μ is defined using small, medium, and large fuzzy distributions. ij (x)(i=1,2,3;j=1,2,3);

[0071] The membership function can be expressed as:

[0072]

[0073] Where: μ i H1, μ i H2, μ i H3 is the sensor v i The measured tool belongs to the three states of initial, normal and acute; x iRepresents the characteristic value actually measured by the sensor measurement system; a i H1, a i H2, a i H3 are the standard characteristic values for the tool to be in these 3 states as measured by sensor i; they can be replaced by the mathematical expectation of the characteristic values measured by sensor v for a certain type of tool state i

[0074] k i H1, k i H2, k i H3 are control coefficients, all constants greater than 0. Obviously, 0 ≤ μ ij ≤ 1. According to μ ij the fuzzy relation matrix R can be obtained, and by performing normalization processing on it, a normalized relation matrix is obtained

[0075] When making a decision rule determination, the probabilities of false alarm rate and missed alarm rate detected by the sensor need to be excluded

[0076] The fuzzy vector A is determined by the weights of the sensors, and the sensor weights a i (i = 1, 2, 3) respectively represent vibration sensor 1, vibration sensor 2, and power sensor 3, and their determination is based on the confidence of each sensor. For the simple detection of tool wear state, binary determination is often used. Assume H 0i and H 1i respectively represent that when using the sensor to monitor the tool, it is judged to be in the set wear state and not in the set wear state. D 0i and D 1i represent selecting the hypothesis H 0i to be true and the hypothesis H 1i to be true. P Fi and P Mi respectively represent the false alarm rate and missed alarm rate detected by sensor i. Then:

[0077] P Fi = P(D 1i / H 0i )

[0078] P Mi = P(D 0i / H 1i )

[0079] The false alarm rates and missed alarm rates of the 3 sensors in the experiment are shown in the following table

[0080] sensor False alarm rate False negative rate Power Sensor 1 0.08 0.06 Power Sensor 2 0.07 0.08 Hall Power Sensor 0.10 0.12

[0081] The weights of the sensors are:

[0082] ​

[0083] The normalized fuzzy vector A=(a1, a2, a3) can be obtained from the sensor weights.

[0084] After determining the fuzzy vector A and the fuzzy relationship matrix R * After that, according to the fuzzy weighted linear transformation:

[0085]

[0086] Then each element B in B j (j=1,2,3), which is the possibility of judging that the tool belongs to state j after the fusion of the information of the three sensors during tool state monitoring. When making decisions on the tool state, the following basic principles are adopted:

[0087] The determined target state should have the largest membership value;

[0088] The membership value of the determined target state must be greater than a certain threshold, which is 0.5 in this paper;

[0089] The difference between the membership values ​​of the determined target state and other states must be greater than a single threshold, which is taken as 0.2 in this paper.

[0090] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0091] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A shield machine tool state analysis and diagnosis system, comprising a signal acquisition module, a processing module and an image processing module, characterized in that: The signal acquisition module includes two vibration sensors and a three-phase Hall power sensor, and the processing module includes an industrial computer, which is connected to the vibration sensor and the three-phase Hall power sensor signals, and performs multi-resolution analysis on the vibration signal through the Mallat fast algorithm, and decomposes and reconstructs the vibration signal at five scales to obtain low-frequency signals and high-frequency signals at five scales as characteristic signals of tool wear; The image processing module includes a high-definition camera, which is connected to an industrial computer by signal. The industrial computer is used to perform clarity processing on the shield machine tool image taken by the high-definition camera.

2. A shield machine tool status analysis and diagnosis system according to claim 1, characterized in that: The industrial computer matches the characteristic signal of tool wear with the shield machine tool image and analyzes the shield machine tool state.

3. A diagnostic method for a shield machine tool status analysis and diagnosis system according to any one of claims 1 to 2, characterized in that: The following steps are involved: S1: The vibration sensor and the three-phase Hall power sensor obtain the vibration data, speed data and input power signal of the shield machine tool and the main shaft drive motor. After the signal is transmitted to the industrial computer, the vibration signal is analyzed by the Mallat fast algorithm at multiple resolutions, and decomposed and reconstructed at five scales to obtain low-frequency and high-frequency signals at five scales as the characteristic signals of tool wear. S2: Two vibration sensors and three-phase Hall power sensors obtain a total of 9 groups of data. Fuzzy set theory uses membership functions to represent the uncertainty of each sensor information. Then, fuzzy transformation is used for data processing. Multi-sensor fuzzy fusion theory is applied to fuse the three groups of membership values. After the three sensors are fused, the membership values ​​of the three states of the tool being measured are obtained. Then, the state of the tool being measured is determined according to the decision rule. S3: The shield machine tool image taken by the high-definition camera is processed for clarity, the characteristic signal of tool wear by the industrial computer is matched with the shield machine tool image, and the shield machine tool status is manually analyzed.

4. A shield machine tool state analysis and diagnosis method and system according to claim 1, characterized in that: The signal acquisition parameters in S1 are: the acquisition frequencies of the vibration signal and the power signal are both 10 kHz.

5. A shield machine tool status analysis and diagnosis method and system according to claim 1, characterized in that: The fusion processing of the membership values ​​in S2 includes the following steps: S2.1: Assume that the tool state is given by u j (j=1,2,3) indicates that the sensor is represented by v i (i=1,2,3) indicates that the sensor v i Inferred to be state u j The probability is r i j, that is, v i To u j The membership degree can be constructed by constructing the fuzzy relationship matrix R = (r i j) m×n ; S2.2: The contribution of each sensor to the tool state decision is called sensor weight, expressed as a i (i=1,2,3) indicates that the fuzzy vector A=(a1,a2,a3) indicates the credibility of the judgment of various sensors; S2.3: The possibility of the tool state obtained by fuzzy transformation B = A·R = (b1, b2, b3), b j (j=1,2,3) is the membership degree R of the fuzzy subset obtained by comprehensive judgment of the tool state.

6. A shield machine tool status analysis and diagnosis method and system according to claim 5, characterized in that: The fuzzy relationship matrix R must determine the form of the membership function, that is, determine the membership function form of the characteristic values ​​observed by each sensor relative to each state of tool wear, and divide the tool wear state into three fuzzy states (initial, normal, and acute).

7. A shield machine tool state analysis and diagnosis method and system according to claim 5, characterized in that: The decision rule also needs to exclude the probability of false alarm rate and missed alarm rate detected by the sensor when determining.

8. A shield machine tool status analysis and diagnosis method and system according to claim 3, characterized in that: The method for clearing the image in S3 is to perform random amplitude enhancement and attenuation after slicing.

Images