Cutter detection device based on pulsed eddy current and detection method thereof

Abstract

The cutter detection device based on the pulsed eddy current is characterized by comprising a coil assembly, the coil assembly comprises an excitation coil (1), a detection coil I (2-1), a detection coil II (2-2) and a detection coil III (2-3); the end face of the excitation coil (1) and the end face of the detection coil I (2-1) face a tool. The detection coil I (2-1) is arranged between the cutter and the excitation coil (1); the detection coil II (2-2) and the detection coil III (2-3) are coaxially and symmetrically arranged on the two sides of the cutter; the excitation coil (1) excites the surface of the tool through square wave pulse signals to generate an eddy current field, and the detection coil is used for capturing eddy current field changes and extracting related signal features. The tool detection method comprises the steps of characteristic quantity collection, sample construction, machine learning model training, precision verification and the like. According to the invention, active detection of the abrasion loss of the cutter is realized, and active online detection of the failure form and the failure degree of the cutter can also be realized.

Description

Technical Field

[0001] The invention relates to a tool detection device and a detection method thereof, and in particular to a tool detection device based on pulse eddy current and a tool wear amount and tool failure form degree detection method. Background Art

[0002] The development and utilization of underground space resources, as well as the mining of underground mineral resources, are inseparable from various types of rock breaking equipment. The rock breaking equipment relies on various rock breaking tools to cut and crush the generalized rock and soil media (including various types of ores, coal rocks and compacted soil) on the face or working surface. Due to the complex and changeable underground geological conditions of the rock breaking equipment, the harsh rock breaking load conditions, and the complex mechanism of the knife-rock action, the cutting edge of the rock breaking tool is very prone to various failure accidents such as excessive wear.

[0003] Taking the annular rolling rock-breaking cutter (hereinafter referred to as the cutter) as an example, this type of cutter gradually penetrates the face rock layer under the combined effect of the vertical rock-invading load of the vertical face and the friction torque around its own cutter shaft, and cuts the face rock layer by means of rotary rolling. At the same time, under the coupled rock-breaking effect of adjacent cutters, the face rock is continuously cut and crushed. The cutting edge of the cutter (hereinafter referred to as the cutting edge) is very prone to normal wear failure (uniform radial wear, and when the radial wear reaches 35mm), chord wear, eccentric wear, fracture and other forms of failure. In particular, under extreme geological conditions such as three highs (high confining pressure, high rock hardness and high quartz content), the time for replacing or repairing the cutter of large underground excavation equipment accounts for about 1 / 3 of the total working time, and the cost of the cutter is about 1 / 3 of the excavation construction cost. It can be seen that under the service conditions of rock breaking equipment, real-time monitoring of the status of rock breaking tools and timely detection and replacement of failed tools can effectively avoid serious accidents such as overall damage to the tools, deformation and disintegration of the cutter disc caused by tool failure, and help reduce the cost of tool replacement, ultimately ensuring the reliable operation of such large-scale underground excavation equipment.

[0004] At present, the engineering community mainly adopts the method of "regular shutdown and manual maintenance" to ensure the construction safety of such large-scale underground excavation equipment. In the field of tool wear inspection, there are passive detection methods such as adding odorants with pungent odors to bearing lubricants, but such methods have poor real-time performance and low feasibility; in addition, there are non-contact tool wear detection methods, but the lifting distance is too short (less than 35mm) and the probe is easily affected by the scratches of rock ballast, which is not suitable for the harsh working conditions of underground excavation. It is worth noting that there is no mature technology to reliably detect the working status of the tool in real time and accurately distinguish the failure form of the tool. Summary of the invention

[0005] In order to overcome the above limitations, the present invention provides a tool detection device based on pulsed eddy current, which is characterized by:

[0006] It includes a coil assembly; the coil assembly includes an excitation coil and a detection coil I; the end faces of the excitation coil and the detection coil I are both arranged facing the tool; the detection coil I is arranged between the tool and the excitation coil, and the detection coil I and the excitation coil are coaxially configured; a square wave pulse signal is input into the excitation coil to stimulate the tool surface to be inspected to generate an eddy current field; the detection coil I captures the changes in the eddy current field, and collects the eddy current signal, and further extracts the characteristic parameters related to the tool wear.

[0007] The detection coil I is arranged at a certain distance from the surface to be inspected of the tool.

[0008] Preferably, the coil assembly also includes a detection coil II and a detection coil III; the detection coil II and the detection coil III are coaxially and symmetrically arranged on both sides of the tool; the detection coil I, the detection coil II and the detection coil III capture the changes in the eddy current field, and further extract characteristic parameters related to the failure form and degree of the tool.

[0009] More preferably, the parameters of the coil assembly are set as follows: the number of coil turns is 200; the inner diameter of the excitation coil is 36 mm and the outer diameter is 70 mm; the inner and outer diameters of the detection coil are 30 mm and 40 mm respectively; the excitation coil height is 11.8 mm and the detection coil height is 10 mm.

[0010] A tool detection method used in conjunction with a tool detection device based on pulsed eddy current of the present invention is characterized in that it comprises the following steps:

[0011] S1: Collect eddy current signals corresponding to tool wear and extract characteristic quantities related to tool wear;

[0012] S2: Construct characteristic quantity samples related to tool wear;

[0013] S3: Establish a tool detection and analysis model based on the machine learning algorithm; S4: Verify the accuracy of the model. If the accuracy requirement is met, proceed to S5, otherwise return to S2 and continue to increase the number of feature samples;

[0014] S5: extracting the feature value of unknown tool;

[0015] S6: Predict tool wear.

[0016] Preferably, in S1, the integral value of the voltage signal is selected as the characteristic value.

[0017] Preferably, in S1, characteristic quantities are collected by physical test means; the tool is cut into annular units with equal radial thickness; and tools with different tool wear amounts are simulated by stacking annular units layer by layer.

[0018] More preferably, the characteristic quantities are collected by finite element simulation; a section is cut from the tool, and a stepped finite element model of the tool with the same interval thickness is established; and the characteristic quantities are continuously collected by parametric scanning.

[0019] More preferably, in S3, a BP neural network or a SVM can be selectively used to establish a tool detection analysis model;

[0020] A tool detection method used in conjunction with a tool detection device based on pulsed eddy current of the present invention is characterized in that it comprises the following steps:

[0021] S1: Determine the failure mode and failure degree of the tool;

[0022] S2: Construct a sample of characteristic quantities related to a given failure form and its given failure degree;

[0023] S3: Training tool detection and analysis model based on machine learning algorithm;

[0024] S4: Verify the model accuracy. If the accuracy requirement is met, go to S5. Otherwise, return to S2 and continue to increase the number of feature samples.

[0025] S5: extracting the feature value of unknown tool;

[0026] S6: Predict the failure mode and degree of tool failure.

[0027] Preferably, in step S1, the failure forms of the tool are divided into fracture, eccentric wear, and chord wear / normal wear failure; the degree of failure of the failure form can be characterized according to the degree characterization variable and the size of the interval gradient. The fracture width ranges from 1 to 5 mm, and the interval gradient is 0.2 mm; the eccentric wear ranges from 1 to 8 mm, and the interval gradient is 1 mm; the chord wear or normal wear ranges from 10 to 30 mm, and the interval gradient is 2 mm.

[0028] Preferably, in step S2, the coil voltage peak value Umax, peak time T and coil flux linkage Ψ are selected as characteristic quantities.

[0029] The advantages of the tool detection device and detection method based on pulsed eddy current of the present invention are:

[0030] 1. It not only realizes the active detection of tool wear, but also realizes the active online detection of tool failure form and failure degree, providing a basis for the construction party to formulate tool inspection and maintenance plan in time.

[0031] 2. The lift-off distance of the tool detection device based on pulsed eddy current is increased by about 70% compared with the traditional eddy current detection method, effectively avoiding the scratching effect of rock chips. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 The present invention is a schematic diagram of the relative arrangement of a coil component of a tool detection device and a tool to be detected.

[0033] Figure 2 The figure is a working principle diagram of a tool detection device of the present invention.

[0034] Figure 3 The present invention is a flowchart of a tool detection method used in conjunction with a tool detection device according to a first specific embodiment of the present invention.

[0035] Figure 4 This is the relationship curve between tool wear d and magnetic flux Ψ.

[0036] Figure 5 Schematic diagram of tool truncation and tool step finite element model establishment.

[0037] Figure 6 Schematic diagram of the simulation of the tool stacking ring elements layer by layer.

[0038] Figure 7 The present invention is a flowchart of a tool detection method used in conjunction with a tool detection device according to a second specific embodiment of the present invention.

[0039] Figure 8 For a tool that has fracture failure and a fracture width of w, a finite element simulation model is established with reference to the layout scheme of the coil assembly in the second specific embodiment of the present invention.

[0040] Fig. 9 For a tool that has eccentric wear failure and an eccentric wear amount of e, a finite element simulation model is established with reference to the layout scheme of the coil assembly in the second specific embodiment of the present invention.

[0041] Fig.10 For a tool that has suffered string grinding failure and a string grinding depth of r, a finite element simulation model is established with reference to the layout scheme of the coil assembly in the second specific embodiment of the present invention.

[0042] Fig.11 This is a curve diagram showing the relationship between the tool string grinding depth r and the coil magnetic flux Ψ.

[0043] Fig.12 This is a curve diagram showing the relationship between the tool string grinding depth r and the coil voltage peak value Umax.

[0044] Fig.13 This is a curve diagram showing the relationship between the tool string grinding depth r and the coil peak time T.

[0045] Fig.14 This is a schematic diagram of the structure of the BP neural network.

[0046] Fig.15 This is a comparison chart of the prediction results of the BP neural network test set.

[0047] Fig.16 Schematic diagram of the classification principle of support vector machine (SVM).

[0048] Fig.17 This is a comparison chart of the prediction results of the SVM training set.

[0049] Fig.18 This is a comparison chart of the prediction results of the SVM test set.

[0050] In the figure:

[0051] 1. Excitation coil; 2-1. Detection coil I; 2-2. Detection coil II; 2-3. Detection coil III; 3. Tool; 4. Stepped finite element model of the tool. DETAILED DESCRIPTION

[0052] Specific embodiment one.

[0053] like Figure 1 As shown, the present invention is a tool detection device based on pulsed eddy current, characterized in that: it includes a coil assembly; the coil assembly includes an excitation coil (1) and a detection coil I (2-1); the end face of the excitation coil (1) is arranged facing the surface to be inspected of the tool to be detected (hereinafter referred to as the tool); in this example, Figure 1 As shown, the tool is an annular rolling rock breaking tool. Therefore, in order to improve the detection accuracy of the present invention, the axis of the excitation coil (1) can be arranged perpendicular to the axis of the tool according to the contour characteristics of the tool, and the axis of the excitation coil (1) is located on the symmetrical middle plane of the tool;

[0054] The detection coil I (2-1) is directly facing the surface to be inspected of the tool to be inspected (hereinafter referred to as the tool), and is arranged between the surface to be inspected and the excitation coil (1), and the detection coil I (2-1) and the excitation coil (1) are coaxially arranged;

[0055] A specific embodiment of a tool detection device based on pulsed eddy current according to the present invention can detect the following Figure 1 The wear amount d of the tool shown is determined by the following principle: Figure 2As shown. In the figure, a square wave pulse signal (001) generated by a signal generating component (not shown) is input into an excitation coil (1), so that the excitation coil (1) generates an alternating magnetic field (003); the alternating magnetic field (003) will excite the surface to be inspected of the tool to generate an eddy current field (004); when the tool fails, such as when the surface to be inspected has defects such as wear, the distribution characteristics of the eddy current field (004) located on the surface to be inspected of the tool will change, such as the eddy current intensity and shape will no longer be uniform; the change of the eddy current field (004) is captured by the detection coil I (2-1), and the eddy current signal (002) is collected and obtained; the characteristic parameters of the eddy current signal (002) are further extracted to realize the detection of the tool wear amount d.

[0056] like Figure 1 As shown, in order to prevent rock chips from scratching and damaging the detection coil I, the detection coil I is set at a certain distance (defined as the lift-off distance L) from the surface to be inspected of the tool (before failure occurs).

[0057] Specific embodiment two.

[0058] Preferably, in order to enable the tool detection device based on pulsed eddy current of the present invention to detect more failure state parameters and realize the classification of tool failure forms and degrees, such as Figure 1 As shown, the coil assembly also includes a detection coil II (2-2) and a detection coil III (2-3); the detection coil II (2-2) and the detection coil III (2-3) are coaxially and symmetrically arranged on both sides of the tool.

[0059] Similarly, if Figure 1 As shown, when the tool fails to varying degrees, the distribution characteristics of the eddy current field (004) located on the surface of the tool to be inspected will change; the changes in the eddy current field (004) are captured by the coil assembly, and the eddy current signal (002) is acquired; and the characteristic parameters of the eddy current signal (002) are further extracted to detect the form and degree of tool failure.

[0060] A tool detection method used in conjunction with a tool detection device based on pulsed eddy current according to a specific embodiment of the present invention is characterized in that: Figure 3 Steps shown:

[0061] S1: acquiring eddy current signals (002) corresponding to different tool wear amounts d, and extracting characteristic quantities related to the tool wear amount d in the eddy current signals (002);

[0062] Preferably, in S1, the characteristic quantity is selected as the integral value of the voltage signal; Figure 4As shown in Figure 2, simulation studies show that as the tool wear d increases, the integral value (magnetic flux) of the voltage signal in the eddy current signal (002) shows an obvious exponential decay trend. This indicates that there is a high correlation between the voltage integral value and the wear amount, which can be used to accurately evaluate the degree of tool wear.

[0063] Typically, physical testing methods can be used to collect the characteristic quantity. Preferably, in order to reduce the test cost (such as the need to produce tool samples with different tool wear amounts d), Figure 5 As shown, the tool is cut into annular units with the same interval thickness Δd by using wire cutting technology; similar to building blocks, the units are stacked layer by layer to simulate tools with different tool wear amounts d; then, with the help of the specific embodiment 1 of the tool detection device based on pulsed eddy current of the present invention, the eddy current signal (002) of the given tool is measured to further extract the characteristic quantity;

[0064] More preferably, considering the limitations of time-consuming, labor-intensive and costly physical test methods to collect the characteristic quantities, finite element simulation methods are used to collect the characteristic quantities. More specifically, for different tool wear amounts d, corresponding finite element simulation models are established respectively; characteristic quantities are obtained through simulation technology;

[0065] More preferably, in order to reduce the time of repeated modeling and further improve the simulation efficiency, such as Figure 6 As shown, a section is cut from the tool, and a step-shaped finite element model of the tool with the same interval thickness Δd is established; further, referring to the layout scheme of the coil assembly in the specific embodiment 1 of the tool detection device based on pulsed eddy current of the present invention, the characteristic quantities corresponding to different tool wear amounts d are continuously collected by parametric scanning. More specifically, COMSOL simulation software is used to simulate and solve the step-shaped finite element model of the tool.

[0066] S2: Construct characteristic quantity samples related to tool wear d;

[0067] S3: Based on the machine learning algorithm, a tool detection and analysis model is established, and the model is trained using some of the feature samples obtained in step S2; more specifically, a BP neural network and a SVM (support vector machine) algorithm can be used, and the tool detection and analysis model is established in a Matlab software environment;

[0068] S4: Use the remaining feature samples obtained in step S2 to verify the accuracy of the tool detection and analysis model; if the accuracy of the tool detection and analysis model meets the requirements, such as not less than 90%, then go to S5, otherwise return to S2 and continue to increase the number of feature samples;

[0069] S5: using the tool detection device based on pulsed eddy current according to the first embodiment of the present invention to detect a tool with a given unknown wear amount, and extracting a characteristic amount;

[0070] S6: Input the feature quantity obtained in step S5 into the tool detection analysis model to measure the wear amount of the tool.

[0071] Taking into account that the parameters of the coil assembly will have a significant impact on the signal strength, sensitivity, detection range, etc., and thus affect the detection accuracy of the tool detection device based on pulsed eddy current of the present invention, preferably, before step S1, the parameters of the coil assembly need to be optimized.

[0072] It is more preferred to use finite element simulation technology to optimize the number of coil turns of the coil assembly, the inner and outer diameters of the coil, and the height of the coil. This is because: the number of coil turns determines the strength of the electromagnetic field and the strength of the induced signal; usually the number of turns of the excitation coil is increased to ensure that the present invention has sufficient magnetic field strength, and the number of turns of the detection coil is increased to optimize the signal induction sensitivity; the inner and outer diameters of the coil determine the coverage of the coil and affect the size and uniformity of the eddy current induction area; the height of the coil affects the distribution of eddy currents and the effectiveness of signal reception. Simulation analysis shows that when the number of coil turns is 200, the magnetic field strength and system efficiency are optimally balanced; the inner diameter of the excitation coil is 36mm and the outer diameter is 70mm, and the eddy current coverage range is now; the inner and outer diameters of the detection coil are 30mm and 40mm respectively, and the induced signal is clear and strong; the height of the excitation coil is 11.8mm, and the height of the detection coil is 10mm, and the electromagnetic field coverage area is sufficient, and the signal transmission and induction effects are optimal.

[0073] A tool detection method used in conjunction with a tool detection device based on pulsed eddy current according to a second specific embodiment of the present invention is characterized in that: Figure 7 Steps shown:

[0074] S1: Determine the failure form and failure degree of the tool; in this example, the failure form of the tool is divided into fracture, eccentric wear, chord wear or normal wear failure (chord wear / normal wear failure); the failure degree of the failure form can be characterized by referring to the corresponding degree characterization variables and the size of the interval gradient in Table 1. For example, the failure degree of the two failure forms of chord wear or normal wear failure is divided into 11 levels of "10, 12, 14, ... 28 and 30mm" according to the size of the degree characterization variable r and its interval gradient Δr, among which r is 30mm, indicating that the chord wear or normal wear failure is the most serious. At present, it is temporarily impossible to effectively further distinguish between chord wear and normal wear failure.

[0075] Table 1 Failure forms of cutting tools and their corresponding failure degrees

[0076]

[0077] S2: Similarly, construct a characteristic quantity sample related to a given failure form and a given failure degree; in this example, similarly, under the COMSOL simulation software, refer to the layout scheme of the coil component in the specific embodiment 2 of the tool detection device based on pulsed eddy current of the present invention, and establish the following Figures 8 to 10 The finite element simulation model shown uses finite element simulation to collect characteristic quantities;

[0078] Preferably, in step S2, the characteristic quantities are selected as the maximum voltage value output by the detection coil (coil voltage peak value Umax), the time when the voltage signal reaches the peak value (peak time T), and the magnitude of the magnetic flux passing through the detection coil (coil magnetic flux Ψ). Figures 11 to 13 As shown in the figure, they are the relationship curves between the tool string grinding depth r and the above three characteristic quantities obtained in the study. Figures 11 to 13 In the figure, the voltage peak and flux signal of coil 1 show a downward trend, while the signals of coils 2 and 3 first rise and then fall or gradually increase. The peak time of coil 1 changes significantly when the string grinding depth reaches 22 mm, indicating that different coils have different response characteristics to the string grinding depth.

[0079] It can be seen from the figure that the above three characteristic quantities can be used to characterize the string grinding depth r of the tool.

[0080] S3: Similarly, based on the machine learning algorithm, a tool detection and analysis model is established, and the model is trained using some of the feature samples obtained in step S2; more specifically, the BP neural network and SVM are also used to establish the tool detection and analysis model in the Matlab software environment;

[0081] S4: Use the remaining feature samples obtained in step S2 to verify the accuracy of the tool detection and analysis model; if the accuracy of the tool detection and analysis model meets the requirements, such as not less than 90%, then go to S5, otherwise return to S2 and continue to increase the number of feature samples;

[0082] S5: using the second specific embodiment of the tool detection device based on pulsed eddy current of the present invention to detect a tool with a given unknown failure form and failure degree, and extracting a characteristic value;

[0083] S6: Input the characteristic quantity obtained in step S5 into the tool detection and analysis model to obtain the failure form and failure degree of the tool.

[0084] In this example, more specifically, if the BP neural network is used to establish a tool detection analysis model to detect the failure form and failure degree of the tool, the specific steps include:

[0085] 1) Input features and network structure:

[0086] like Fig.14 As shown in Figure 1, the basic structure of the input layer, hidden layer and output layer is shown. The input layer receives the characteristic signal of tool detection, such as voltage peak x1, x2, ..., x d , these input features are passed to the hidden layer through weighting.

[0087] The weighted summation formula from the input layer to the hidden layer is:

[0088]

[0089] Among them, v ih Represents the weight between the input layer and the hidden layer. The output of each neuron in the hidden layer is processed by the activation function to capture the complex nonlinear relationship between the input features.

[0090] 2) Forward propagation calculation:

[0091] The hidden layer passes the weighted signal to the output layer, and the output layer nodes y1, y2, ..., y j Corresponding to different failure types (such as breakage, string wear, and eccentric wear).

[0092] The weighted summation formula from the hidden layer to the output layer is:

[0093]

[0094] Among them, ω hj is the weight between the hidden layer and the output layer. The output layer neurons give the classification results for each failure type based on these weighted sums.

[0095] 3) Output and failure type classification:

[0096] Each node in the output layer represents the probability of a different failure type. As shown in the figure, the y1, y2, ..., y j They correspond to different failure forms, such as tool breakage, chord wear or eccentric wear. By comparing the output values, the network will select the output node with the highest probability as the predicted failure type.

[0097] 4) Model verification and testing:

[0098] like Fig.15 As shown in Figure 2, the predicted value (circle) of the tool detection analysis model is almost consistent with the true value (asterisk), and the mean square error (MSE) is 0.00026615, which indicates that the prediction error of the model is extremely small, and the determination coefficient (R 2 =0.99966) is close to 1, indicating that the model fits the data very well. This verifies that the BP neural network has good generalization ability and high accuracy on the test set.

[0099] In this example, more specifically, if SVM is used to establish a tool detection analysis model to detect the failure form and degree of tool failure, the specific steps include:

[0100] 1) Data preprocessing

[0101] Since the magnitude of different features may be different, SVM is sensitive to the size of the feature value, so the feature data needs to be standardized to ensure that the feature quantities are in the same range. The processed data set is divided into a training set and a test set. The training set is used to build the model, and the test set is used to verify the effect of the model.

[0102] 2) Model training (construction of SVM hyperplane)

[0103] The SVM algorithm is used to find a hyperplane in the feature space that can optimally separate different failure forms. A hyperplane is a decision boundary used to distinguish different categories, and a support vector is the data point closest to the hyperplane. The goal of SVM is to find a hyperplane that maximizes the interval between data points of different categories on both sides of the classification boundary, thereby improving the generalization ability of the model. Fig.16 As shown, the construction process of SVM hyperplane and support vector is demonstrated.

[0104] 3) Select kernel function and parameter optimization

[0105] For linearly separable data, a linear kernel function can be selected. For complex nonlinear data, a nonlinear kernel function is usually used to map the data to a high-dimensional space for classification. Through methods such as cross-validation or grid search, the hyperparameters in the SVM are adjusted to improve the classification effect of the model.

[0106] 4) Model training results and verification

[0107] The training data is input into the model, and the model outputs the predicted failure mode. Fig.17 The figure shows the comparison between the actual classification and the predicted classification of the training set. Mean square error (MSE) and determination coefficient R 2 The value of indicates that the training accuracy of the model is very high. The model is verified by the test set data to observe the performance of the model on unseen data. Fig.18 As shown, the comparison results of the actual value and predicted value of the test set, MSE and R 2 It also shows that the model performs well on the test set.

[0108] 5) Result analysis and optimization

[0109] The mean square error (MSE) and the coefficient of determination R 2 Evaluate the performance of the model on the training and test sets. Fig.17 and Fig.18 R 2 =0.99934 indicates that the classification accuracy of the constructed model is very high and the error is very small. According to the classification results and errors, the model can be further optimized by adjusting hyperparameters or selecting other kernel functions to ensure its robustness and generalization ability on different data sets.

[0110] In the tool failure form detection, BP neural network and SVM both show good performance. Both methods have their own advantages and disadvantages. BP neural network is suitable for nonlinear and large data sets, while SVM is more suitable for small samples and linear or near-linear problems.

[0111] In combination with the above specific embodiments, it can be seen that the tool detection device and detection method based on pulsed eddy current of the present invention are beneficial in that:

[0112] 1. It not only realizes the active detection of tool wear, but also realizes the active online detection of tool failure form and failure degree, providing a basis for the construction party to formulate tool repair and maintenance plans in a timely manner. In this way, it can effectively avoid serious accidents such as overall damage of the tool and deformation and disintegration of the cutter disc caused by tool failure, and help reduce the cost of tool replacement, and ultimately ensure the reliable operation of such large underground excavation equipment.

[0113] 2. Experimental results show that the lift-off distance L in the tool detection device based on pulse eddy current is as high as 61 mm, which is about 70% higher than the lift-off distance that can be achieved by the traditional eddy current detection method, effectively avoiding the scratching effect of rock chips.

Claims

1. A tool detection device based on pulsed eddy current, characterized in that: The invention comprises a coil assembly; the coil assembly comprises an excitation coil (1) and a detection coil I (2-1); the excitation coil (1) and the detection coil I (2-1) are both arranged facing the tool; the detection coil I (2-1) is arranged between the tool and the excitation coil (1), and the detection coil I (2-1) and the excitation coil (1) are coaxially arranged; a square wave pulse signal is input into the excitation coil (1) to excite the tool surface to be inspected to generate an eddy current field; the detection coil I (2-1) captures the change of the eddy current field, collects the eddy current signal, and further extracts the characteristic parameters related to the tool wear; the detection coil I (2-1) is arranged at a certain distance from the tool surface to be inspected.

2. A tool detection device based on pulsed eddy current according to claim 1, characterized in that: The coil assembly also includes a detection coil II (2-2) and a detection coil III (2-3); the detection coil II (2-2) and the detection coil III (2-3) are coaxially and symmetrically arranged on both sides of the tool; the detection coil I (2-1), the detection coil II (2-2) and the detection coil III (2-3) capture the changes in the eddy current field and further extract characteristic parameters related to the failure form and failure degree of the tool.

3. The tool detection device based on pulsed eddy current according to claim 1, characterized in that: The parameters of the coil assembly are set as follows: the number of coil turns is 200; the inner diameter of the excitation coil (1) is 36 mm and the outer diameter is 70 mm; the inner and outer diameters of the detection coil I (2-1) are 30 mm and 40 mm respectively; the height of the excitation coil (1) is 11.8 mm, and the height of the detection coil I (2-1) is 10 mm.

4. A tool detection method used in conjunction with a tool detection device based on pulsed eddy current as claimed in any one of claims 1 and 3, characterized in that: The steps include: S1: Collect eddy current signals corresponding to tool wear and extract characteristic quantities related to tool wear; S2: Construct characteristic quantity samples related to tool wear; S3: Establish tool detection and analysis model based on machine learning algorithm; S4: Verify the model accuracy. If the accuracy requirement is met, go to S5. Otherwise, return to S2 and continue to increase the number of feature samples. S5: extracting the feature value of unknown tool; S6: Predict tool wear.

5. A tool detection method according to claim 4, characterized in that: In S1, the integral value of the voltage signal is selected as the characteristic quantity; in S3, a BP neural network or SVM can be selectively used to establish a tool detection analysis model.

6. A tool detection method according to claim 4, characterized in that: The characteristic quantities are collected by physical test means; the tool is cut into annular units with equal radial thickness; and the tools with different tool wear amounts are simulated by stacking annular units layer by layer.

7. A tool detection method according to claim 4, characterized in that: The characteristic quantities are collected by finite element simulation. A section is cut from the tool, and a stepped finite element model of the tool with the same interval thickness is established. The characteristic quantities are continuously collected by parametric scanning.

8. A tool detection method used in conjunction with the tool detection device based on pulsed eddy current according to claim 2, characterized in that: The steps include: S1: Determine the failure mode and failure degree of the tool; S2: construct feature sample; S3: training tool detection analysis model; S4: Verify the model accuracy. If the accuracy requirement is met, go to S5. Otherwise, return to S2 and continue to increase the number of feature samples. S5: extracting the feature quantity of unknown tool; S6: Predict the failure mode and degree of tool failure.

9. A tool detection method according to claim 8, characterized in that: In S1, the failure modes of the tool are divided into fracture, eccentric wear, and chord wear / normal wear failure; Table 1 Failure forms of cutting tools and their corresponding failure degrees The failure degree of the failure form can be characterized by referring to the corresponding degree characterization variables and the size of the interval gradient in Table 1.

10. A tool detection method according to claim 8, characterized in that: In S2, the coil voltage peak value Umax, peak time T and coil flux linkage Ψ are selected as characteristic quantities.

Images