A method and system for detecting defects on a drum

Abstract

This invention relates to the field of crack identification technology, and provides a method and system for detecting defects in a roller. The roller defect detection method includes the following steps: S1, obtaining the simulated strain field corresponding to different spatial positions of the roller under different working conditions, and calculating the simulated working condition modal coefficients based on the simulated strain field; S2, setting the set of positions for arranging strain sensors on the roller, and calculating the local true modal coefficients corresponding to the strain sensors based on the simulated working condition modal coefficients; S3, retrieving the optimal modal coefficients based on the local true modal coefficients, and reconstructing the complete strain field based on the optimal modal coefficients; S4, constructing a multi-objective function, determining the sensitive monitoring area, and obtaining the arrangement positions of the strain sensors.

Description

Technical Field

[0001] This invention belongs to the field of crack identification technology, and in particular relates to a method and system for detecting defects in rollers. Background Technology

[0002] Currently, belt conveyors, as core transportation equipment in coal, port, and power industries, are considered one of the three major industrial transportation tools alongside automobiles and trains. They support continuous production in modern industry due to their advantages of large capacity, long distance, and low energy consumption. The rollers, as the core transmission components of the conveyor, are formed by welding sheet metal rolls and flanges. They have longitudinal and circumferential welds and are subjected to heavy loads and unbalanced loads for extended periods, making them highly susceptible to cracking failures. Circumferential weld damage accounts for over 90% of these failures. Microcracks, as early failure characteristics, are characterized by no obvious signs, small size, and high depth-to-width ratio, making them difficult to identify with conventional dye penetrant testing and easily overlooked. If these defects are not detected in time, they can rapidly expand, leading to weld breakage in the roller, belt tearing, and even major accidents such as fires, causing huge economic losses and casualties.

[0003] Existing detection methods mainly include vibration monitoring, acoustic emission monitoring, infrared thermal imaging, ultrasonic or magnetic particle testing, and motor current or drive power detection. Vibration monitoring uses vibration sensors to monitor changes in the roller's motion and identifies faults based on vibration spectrum characteristics. While vibration amplitude changes may occur when the roller undergoes macroscopic structural changes or cracks extend to a large area, vibration signals are difficult to discern when cracks are in their initiation stage due to minimal structural deformation. Acoustic emission monitoring utilizes transient elastic wave signals generated during material crack propagation or fracture. Although acoustic emission can detect crack propagation events, its sensitivity to continuous, stable, and micro-crack growth is limited, and it is easily affected by ambient noise. Infrared thermal imaging infers structural defects by detecting surface temperature anomalies. However, thermal imaging primarily focuses on surface temperature changes and cannot accurately assess internal distribution or micro-cracks, and long-term online monitoring is difficult. Ultrasonic or magnetic particle testing is mostly used for periodic manual inspections and cannot meet the real-time monitoring needs of continuous operation. Furthermore, it is complex to operate, has a low inspection frequency, and may miss some cracks. Motor current or drive power detection methods infer faults by detecting changes in system operating parameters. They are only applicable to stages of large structural damage or severe operational abnormalities and are not sensitive to microcrack detection.

[0004] The common problems with existing roller inspection technologies are: the signal at the microcrack stage is extremely weak and difficult to detect; the sensors used to monitor roller faults are mostly located at single points, resulting in inaccurate monitoring and often requiring the machine to be stopped to detect the fault location, making online monitoring difficult; in general, existing detection methods cannot achieve real-time online monitoring of roller microcracks, let alone identify and warn of crack initiation stages.

[0005] To solve the above-mentioned technical problems, this invention designs a method and system for detecting defects in rollers. Summary of the Invention

[0006] To address the above problems, the present invention provides the following technical solution: a method for detecting defects in a roller, comprising the following steps: S1, obtain the simulated strain field corresponding to different spatial positions of the roller under different working conditions, and calculate the modal coefficient of the simulated working condition based on the simulated strain field; S2, Set the set of positions for the strain sensors arranged on the roller, and calculate the local true modal coefficients corresponding to the strain sensors based on the modal coefficients of the simulation working condition; S3, the optimal modal coefficients are derived from the local true modal coefficients, and the complete strain field is reconstructed based on the optimal modal coefficients; S4. Construct a multi-objective function to determine the sensitive monitoring area and obtain the location of the strain sensor.

[0007] Based on the above technical solution, step S1 includes the following steps: S1.1, obtain the simulated strain field corresponding to different spatial positions of the roller under different working conditions, construct a strain snapshot matrix based on the simulated strain field, and perform mean removal processing on the strain snapshot matrix; S1.2, Perform singular value decomposition on the processed strain snapshot matrix, extract the dominant modes of the strain snapshot matrix, and calculate the low-dimensional orthogonal mode basis matrix corresponding to the dominant modes; S1.3 Calculate the modal coefficients of the simulated working condition based on the simulated strain field and the low-dimensional orthogonal modal basis matrix.

[0008] Based on the above technical solution, step S1.1 includes the following steps: S1.1.1 Discretize the finite element model of the roller into N spatial nodes and obtain M sets of simulated strain fields under different working conditions, including load, torque, and crack state. S1.1.2, construct the strain snapshot matrix based on the simulated strain fields of M groups. The strain snapshot matrix is:

[0009] Where N is the number of spatial points and M is the number of working conditions; S1.1.3, calculate the mean of the simulated strain fields of group M, and remove the mean from the strain snapshot matrix. The calculation formula is as follows:

[0010]

[0011] in, , It is an identity matrix.

[0012] Based on the above technical solution, step S1.2 includes the following steps: S1.2.1, Perform singular value decomposition on the processed strain snapshot matrix. The decomposition formula is as follows:

[0013] in, Let be a left singular matrix, representing the strain space modes; It is a diagonal matrix, and the diagonal elements are singular values, representing the magnitude of the modal energy. Let be the transpose of the right singular matrix, and let represent the transpose of the working condition coefficient matrix. This represents the energy magnitude of the k-th mode. For the k-th strain space mode, This represents the variation trend of the k-th mode under different operating conditions; S1.2.2, extract the first r dominant modes, based on the left singular matrix. Obtain the low-dimensional orthogonal mode basis matrix corresponding to the first r dominant modes. , where r is the number of truncated modes.

[0014] Based on the above technical solution, step S1.3 includes the following steps: S1.3.1, Set the modal coefficients for the simulation conditions. for

[0015] The simulated strain field is represented based on the modal coefficients of the simulated operating conditions.

[0016] Retaining the first r dominant modes, the simulated strain field is expressed as:

[0017] S1.3.2, based on the simulated strain field and low-dimensional orthogonal modal basis matrix Calculate the modal coefficients of the simulation operating conditions

[0018]

[0019] in,() + This represents the inverse operation of a matrix.

[0020] Based on the above technical solution, step S2 includes the following steps: S2.1, Assume p strain sensors are arranged on the roller, and the set of positions of the strain sensors is as follows:

[0021] in, ; S2.2, Define the measurement point selection matrix as follows:

[0022] in

[0023] in, This indicates that the i-th sensor is located at the j-th spatial node; S2.3, Based on the measurement point selection matrix, the actual monitored strain vector is:

[0024] in, Let n be the strain values ​​actually monitored by p sensors, and n be the noise, which can be ignored; combined with the low-dimensional orthogonal modal basis matrix obtained in step S1.2 The strain vector is expressed as

[0025] Based on the low-dimensional orthogonal modal basis matrix Calculate the local true modal coefficients corresponding to the actual monitoring of p strain sensors. .

[0026] Based on the above technical solution, step S3 includes the following steps: S3.1, Given the strain vector y actually measured by the strain sensor, the local true modal coefficients are inverted using the least squares method. The optimal modal coefficients are obtained:

[0027] S3.2, based on the optimal modal coefficients derived from the inversion. Reconstruct the complete strain field as

[0028] Based on the above technical solution, step S4 includes the following steps: S4.1, Based on the measurement point selection matrix in step S2.3 and the low-dimensional orthogonal modal basis matrix in step S1.2, the measurement matrix is ​​defined as follows:

[0029] S4.2, Low-dimensional orthogonal modal basis matrix based on strain sensor measurement points Whether the sensitivity is balanced by setting condition number index function

[0030]

[0031] in, and These represent the maximum and minimum singular values ​​of the measurement matrix H; the reconstruction error function is set based on the overall deviation of the reconstructed strain field from the true strain field.

[0032]

[0033] in, To simulate the real strain field of the working condition, The reconstructed strain field is obtained based on the set S of p strain sensor locations; the modal coefficient error function is set based on the modal error.

[0034] in, The corresponding simulation modal coefficients under the simulation conditions. , The optimal modal coefficients derived from the strain sensor Constructing a multi-objective function

[0035] S4.3 Based on the multi-objective function, the sensitive monitoring area is determined by the multi-objective particle swarm optimization algorithm, and the sensor placement location is obtained.

[0036] Based on the above technical solution, after step S4, the method further includes: S5, calculating the energy ratio of the optimal modal coefficients, the modal coefficient offset, and the strain gradient to determine whether the roller has cracks and to locate the crack positions; step S5 includes the following steps: S5.1, calculate the proportion of the optimal modal coefficient relative to the sum of the optimal modal coefficients corresponding to the first r dominant modes, and use this as the energy proportion to determine whether a crack has occurred. The energy proportion is...

[0037] in, For the k-th optimal modal coefficient, The i-th optimal modal coefficient; S5.2, Calculate the modal coefficient offset. Based on the modal coefficient offset, determine whether the roller deviates from a healthy state. The modal coefficient offset is...

[0038] in, These are the modal coefficients obtained by inversion under the reference state; S5.3, Calculate the strain gradient and locate the crack position based on the strain gradient. The strain gradient is...

[0039] In a second aspect, the present invention provides a roller defect detection system, including a processor and a memory storing program instructions, wherein the processor is configured to execute the roller defect detection method as described in any of the above embodiments when the program instructions are executed.

[0040] Compared with related technologies, the beneficial effects of the present invention are as follows: This invention first performs finite element simulation on the drum to obtain full-field strain data. Then, it establishes a low-dimensional physical subspace of the drum strain field based on intrinsic orthogonal decomposition (POD). Subsequently, it combines the missing data clustering intrinsic orthogonal decomposition method (Gappy C-POD) to reconstruct the complete strain field from a small number of measurement point data. On this basis, the multi-objective particle swarm optimization algorithm (MOPSO) is used to determine the sensor placement location to obtain the optimal monitoring position. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only one embodiment of the present invention. For those skilled in the art, other embodiments can be derived from the provided drawings without creative effort.

[0042] Figure 1 This is a flowchart of the roller defect detection method provided by the present invention; Figure 2 This is a schematic diagram of the structure of the roller provided by the present invention; Figure 3 This is a schematic diagram of the simulated strain field of the roller provided by the present invention; Figure 4 This is a schematic diagram of the reconstructed strain field of the roller provided by the present invention; Figure 5 This is a coordinate schematic diagram of the sensor arrangement position of the roller provided by the present invention; Figure 6 This is a schematic diagram of the crack location of the roller provided by the present invention.

[0043] In the diagram: 1. Shell; 2. Connecting plate; 3. Weld. Detailed Implementation

[0044] The present invention will be further described below with reference to the accompanying drawings and examples: Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0045] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "joining" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0046] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0047] Combination Figure 1 As shown, this disclosure provides a method for detecting defects in a roller, including the following steps: S1, obtain the simulated strain field corresponding to different spatial positions of the roller under different working conditions, and calculate the modal coefficient of the simulated working condition based on the simulated strain field; S2, Set the set of positions for the strain sensors arranged on the roller, and calculate the local true modal coefficients corresponding to the strain sensors based on the modal coefficients of the simulation working condition; S3, the optimal modal coefficients are derived from the local true modal coefficients, and the complete strain field is reconstructed based on the optimal modal coefficients; S4. Construct a multi-objective function to determine the sensitive monitoring area and obtain the location of the strain sensor.

[0048] The roller defect detection method provided in this embodiment first performs finite element simulation on the roller to obtain full-field strain data. Then, a low-dimensional physical subspace of the roller strain field is established based on intrinsic orthogonal decomposition (POD). Subsequently, the missing data clustering intrinsic orthogonal decomposition method (Gappy C-POD) is combined to reconstruct the complete strain field from a small number of measurement point data. On this basis, the multi-objective particle swarm optimization algorithm (MOPSO) is used to determine the sensor placement location to obtain the optimal monitoring location.

[0049] Based on the above technical solution, step S1 includes the following steps: S1.1, obtain the simulated strain field corresponding to different spatial positions of the roller under different working conditions, construct a strain snapshot matrix based on the simulated strain field, and perform mean removal processing on the strain snapshot matrix; S1.2, Perform singular value decomposition on the processed strain snapshot matrix, extract the dominant modes of the strain snapshot matrix, and calculate the low-dimensional orthogonal mode basis matrix corresponding to the dominant modes; S1.3 Calculate the modal coefficients of the simulated working condition based on the simulated strain field and the low-dimensional orthogonal modal basis matrix.

[0050] Further, step S1.1 includes the following steps: S1.1.1 Discretize the finite element model of the roller into N spatial nodes and obtain M sets of simulated strain fields under different working conditions, including load, torque, and crack state. S1.1.2, construct the strain snapshot matrix based on the simulated strain fields of M groups. The strain snapshot matrix is:

[0051] Where N is the number of spatial points and M is the number of working conditions; Specifically, For the j-th working condition, the entire drum strain field is represented by the strain snapshot matrix. Each row in the matrix corresponds to a spatial location, and each column represents a working condition.

[0052] S1.1.3, calculate the mean of the simulated strain fields of group M, and remove the mean from the strain snapshot matrix. The calculation formula is as follows:

[0053]

[0054] in, , It is an identity matrix.

[0055] Specifically, in step S1.1.1, the simulated strain fields under different working conditions are obtained through simulation software. The corresponding loads, torques, and crack states under different working conditions are input, and the corresponding simulated strain fields are output. The crack states are preset manually. The main function of step S1.1 is to eliminate static bias, ensuring that the subsequent POD mode description is of deformation modes and not non-constant terms.

[0056] Further, step S1.2 includes the following steps: S1.2.1, Perform singular value decomposition on the processed strain snapshot matrix. The decomposition formula is as follows:

[0057] in, Let be a left singular matrix, representing the strain space modes; It is a diagonal matrix, and the diagonal elements are singular values, representing the magnitude of the modal energy. Let be the transpose of the right singular matrix, and let represent the transpose of the working condition coefficient matrix. This represents the energy magnitude of the k-th mode. For the k-th strain space mode, This represents the variation trend of the k-th mode under different operating conditions; S1.2.2, extract the first r dominant modes, based on the left singular matrix. Obtain the low-dimensional orthogonal mode basis matrix corresponding to the first r dominant modes. , where r is the number of truncated modes.

[0058] Specifically, the first r dominant modes are extracted using POD to capture the core deformation characteristics. Here, r is determined based on the degree of influence of different working conditions on the roller cracking. For example, a larger load has a greater impact on the roller, and a larger torque has a greater impact on the roller. The first r dominant modes are determined in this way.

[0059] Further, step S1.3 includes the following steps: S1.3.1, Set the modal coefficients for the simulation conditions. for

[0060] The simulated strain field is represented based on the modal coefficients of the simulated operating conditions.

[0061] Retaining the first r dominant modes, the simulated strain field is expressed as:

[0062] S1.3.2, based on the simulated strain field and low-dimensional orthogonal modal basis matrix Calculate the modal coefficients of the simulation operating conditions

[0063]

[0064] in,() + This represents the inverse operation of a matrix.

[0065] Further, step S2 includes the following steps: S2.1, p strain sensors are arranged on the roller, and the set of positions of the strain sensors is as follows:

[0066] in, ; S2.2, Define the measurement point selection matrix as follows:

[0067] in

[0068] The measurement point selection matrix is ​​used to extract the measured values ​​of the corresponding sensor locations from the full-field strain. Where N is the total number of nodes on the roller, p is the number of sensors, and n... i This represents the node number where the i-th sensor is located. C i，j =1 indicates that the i-th sensor is positioned at the j-th spatial node; otherwise, it is 0. Each row of the measurement point selection matrix corresponds to a sensor, and each column corresponds to a spatial node. The sensor placement positions are determined by an optimization algorithm to minimize the strain reconstruction error based on a small number of measurement points while ensuring the identifiability of modal information. Specifically, a multi-objective optimization model is constructed with the objectives of minimizing strain reconstruction error and maximizing modal independence, and the optimal set of measurement point positions S is determined using the Multi-Objective Particle Swarm Optimization (MOPSO) algorithm.

[0069] S2.3, based on the measurement point selection matrix, the strain vector actually monitored by p strain sensors is:

[0070] in, Let n be the strain values ​​actually monitored by p sensors, and n be the noise, which can be ignored; combined with the low-dimensional orthogonal modal basis matrix obtained in step S1.2 The strain vector is expressed as

[0071] Based on the low-dimensional orthogonal modal basis matrix Calculate the local true modal coefficients corresponding to the actual monitoring of p strain sensors. Where y is the strain actually measured by the sensor, which is a known quantity.

[0072] Further, step S3 includes the following steps: S3.1, Given the strain vector y actually measured by the strain sensor, the local true modal coefficients are inverted using the least squares method. The optimal modal coefficients are obtained:

[0073] The analytical solution to this optimization problem is:

[0074] The physical meaning of the above formula is that, within the measurement point space of p strain sensors, the strain value predicted by the model is... The optimal modal coefficients are obtained by minimizing the error between the measured strain vector y and the actual strain vector. .

[0075] S3.2, based on the optimal modal coefficients derived from the inversion. The reconstructed complete strain field is as follows:

[0076] Specifically, in steps S2 and S3, the Gappy C-POD method is used to obtain the corresponding modal coefficients based on the observation data of the strain sensor at the measuring point through the least squares criterion; then, the optimal modal coefficients are used... With low-dimensional orthogonal modal basis matrix By performing linear combination, the full-field strain distribution can be reconstructed, thus achieving high-precision recovery of the drum strain field with only p strain sensors.

[0077] Further, step S4 includes the following steps: S4.1, Define the measurement matrix According to the measurement point selection matrix in step S2.3 Where p is the number of sensors and n is the total degrees of freedom; and the low-dimensional orthogonal modal basis matrix in step S1.2 is... The measurement matrix is ​​defined as follows:

[0078] S4.2, Constructing a multi-objective optimization function Low-dimensional orthogonal modal basis matrix based on strain sensor measurement points Whether the sensitivity is balanced by setting condition number index function

[0079]

[0080] in, and These are the maximum and minimum singular values ​​of the measurement matrix H, representing its singular value spectrum distribution characteristics. Their ratio (condition number) reflects the system's sensitivity to error amplification.

[0081] The reconstruction error function is set based on the overall deviation of the reconstructed strain field from the true strain field.

[0082]

[0083] in, For the true strain field of the simulated working condition , The reconstructed strain field obtained by Gappy C-POD based on a set of p strain sensor locations S is used to characterize the full-field reconstruction accuracy of the strain sensor placement scheme. The modal coefficient error function is set based on the modal error.

[0084] in, The corresponding simulation modal coefficients under the simulation conditions. , The optimal modal coefficients derived from the strain sensor Constructing a multi-objective function

[0085] S4.3 Based on the multi-objective function, the sensitive monitoring area is determined by the multi-objective particle swarm optimization algorithm, and the sensor placement location is obtained.

[0086] Specifically, step S4.3 includes the following steps: S4.3.1, the sensor layout scheme is represented as a set of measuring points S = {s1, s2, s3, ... s}. p}, where each measuring point s i Candidate locations corresponding to key areas of the roller; the entire set of candidate measurement points is obtained by pre-discretizing the structural surface, which serves as the optimization search space; S4.3.2, a multi-objective particle swarm optimization algorithm is used to encode the layout scheme, with each particle representing a sensor layout combination; the position (measuring point combination) and velocity of the particles are randomly initialized, and algorithm parameters such as particle swarm size and maximum number of iterations are set; S4.3.3, for each particle's corresponding arrangement scheme S, perform the following calculations: construct the measurement point selection matrix C, and calculate the measurement point matrix. ; Calculate the condition number index J1(S); Invert the modal coefficients and reconstruct the strain field based on the Gappy C-POD method, and calculate the reconstruction error J2(S); Calculate the modal coefficient error J3(S). S4.3.4, based on the Pareto dominance relation of multi-objective optimization, performs non-dominated sorting on the particle swarm, selects the current non-dominated solutions, and constructs an external archive, namely the Pareto solution set, to store the historical optimal layout schemes.

[0087] S4.3.5 guides the particle to update its position and velocity based on its individual optimal position and the global optimal solution in the external archive; it repeatedly executes the objective function calculation and Pareto selection process to achieve iterative optimization of the sensor layout scheme.

[0088] S4.3.6 When the maximum number of iterations is reached or the optimization result converges, output the Pareto optimal solution set; combine engineering requirements (such as sensor quantity limit, installation feasibility, etc.) to select the final sensor layout scheme from the Pareto solution set.

[0089] The final sensor arrangement scheme, while ensuring high reconstruction accuracy, improves the numerical stability and noise resistance of the system, and realizes the recovery of full-field strain information under conditions with fewer measurement points.

[0090] Based on the above technical solution, the following steps are included after step S4: S5, calculate the energy ratio, modal coefficient offset and strain gradient of the optimal modal coefficient, and determine whether the roller has cracks and locate the crack location; Step S5 includes the following steps: S5.1, calculate the proportion of the optimal modal coefficient relative to the sum of the optimal modal coefficients corresponding to the first r dominant modes, and use this as the energy proportion to determine whether a crack has occurred. The energy proportion is...

[0091] in, For the k-th optimal modal coefficient, The i-th optimal modal coefficient; A sudden increase in the energy percentage indicates the appearance of a crack; small cracks have relatively small energy abrupt changes. Specifically, it is necessary to determine whether the strain field structure has changed, that is, to see whether the current deformation is overall deformation or local anomaly (crack). Let E2 be the energy percentage of the k-th mode. Under normal conditions, the strain field is smooth and low-order modes dominate. When a crack appears, the strain changes abruptly and cannot be described by low-order modes. By introducing higher-order modes, the energy of the higher-order modes increases, which can be used to determine whether an anomaly has occurred. For example, under normal conditions, E2≈0.05, but at this time, E2≈0.22, indicating that the higher-order modes have increased and a crack is suspected to have appeared.

[0092] S5.2, Calculate the modal coefficient offset. Based on the modal coefficient offset, determine whether the roller deviates from a healthy state. The modal coefficient offset is...

[0093] in, These are the modal coefficients obtained by inversion under the reference state. The modal coefficients are the average of the first r modal coefficients. Since cracks can change the stiffness of the roller structure, the modal coefficients are changed. Therefore, the roller is judged to be damaged based on the modal coefficients. At the same time as the energy ratio in step S7.1, it is judged whether the roller deviates from the healthy state and produces cracks. For example, if D < 0.2 in the healthy state, then D = 0.78, and the roller is judged to be in an abnormal state.

[0094] S5.3, Calculate the strain gradient

[0095] The location of the crack is determined by the strain gradient.

[0096] Specifically, a crack is essentially a sudden change in strain. When the strain is smooth, the gradient is small; when the strain changes abruptly, the gradient is large. Therefore, the strain location is at the peak of the gradient, which means the crack location is at the peak of the strain gradient. For example, G(X1-2)=0.16, G(X2-3)=0.28, G(X3-4)=0.39, G(X4-5)=0.27. Comparing the strain gradient values, the gradient between position 3 and position 4 is the largest, so the crack location is within this region.

[0097] By judging the energy ratio, modal coefficient offset, and strain gradient, and simultaneously determining whether the roller has cracks, errors in judging a single parameter are avoided, thereby reducing recognition errors and improving recognition accuracy.

[0098] By using modal coefficient changes and strain gradient anomalies, microcracks in the roller can be identified and located. The local microstrain changes caused by microcracks on the roller can be monitored in real time, thereby enabling crack identification and determination of crack location.

[0099] To demonstrate the feasibility of the roller defect detection method provided in this embodiment, experiments were conducted, such as... Figure 2 As shown, an artificial opening is made at the weld 3 between the drum skin 1 and the receiving plate 2. The opening shape can be triangular, rectangular, or irregular to simulate the crack state under different working conditions. The frictional force and average pressure on the drum are obtained through simulation, as shown in Table 1, to obtain the simulated strain field. Figure 3 As shown.

[0100] Table 1. Force values ​​per 10° arc surface

[0101] Assuming eight strain sensors are arranged on the drum, the complete strain field is reconstructed based on the simulated strain field and the strain vectors actually monitored by the strain sensors, such as... Figure 4 As shown, comparing the reconstructed strain field with the real strain field reveals that the reconstructed strain field is highly consistent with the simulated strain field, indicating that the method proposed in this invention can achieve high-precision full-field reconstruction under limited measurement points, thereby reducing sensor costs.

[0102] A condition number index function is constructed based on the measurement matrix; a reconstruction error function is constructed based on the simulated strain field and the reconstructed strain field; a modal coefficient error function is constructed based on the simulated modal coefficients and the optimal modal coefficients; a multi-objective function is constructed; and the placement positions of the eight strain sensors are determined using a multi-objective particle swarm optimization algorithm. The coordinates of the placement positions of the eight strain sensors are displayed on [the graph / table]. Figure 5 middle.

[0103] The presence of cracks is determined based on the calculated energy percentage and modal coefficient offset. If cracks are confirmed to have occurred, the crack location is determined based on the calculated strain gradient. Figure 6 As shown, Figure 6 The red mark indicates the location of the crack.

[0104] This disclosure provides a roller defect detection system, including a bridge measurement module, a signal conditioning module, a data acquisition module, a data processing unit, a threshold judgment module, and a display and alarm module. The bridge measurement module consists of a strain sensor, shielded connecting wires, and terminals. The strain sensor is pre-attached to an elastic substrate, and the arc-shaped ends of the strain monitoring component are welded to the surface of the drum. When a crack occurs, the minute strain signal is amplified through this substrate, thereby improving the accuracy of strain gauge monitoring.

[0105] The data acquisition module, using a wireless strain gauge, employs a multi-channel acquisition circuit to simultaneously sample multiple sensors. Its main function is to convert conditioned analog signals into digital signals and acquire them synchronously across multiple channels. This enables multi-point status monitoring of the rollers, and the sampling frequency can be set according to the operating conditions of the corresponding conveyor.

[0106] The data processing unit consists of an embedded processor (MCU / DSP / ARM), memory (RAM, Flash), and digital signal processing algorithm modules. Its main function is to perform noise reduction, filtering, and feature extraction on the acquired strain signals, analyze the stress state changes of the drum at different operating stages, and construct characteristic parameters of the drum's health status.

[0107] The threshold judgment module includes a threshold storage unit (normal threshold, warning threshold, danger threshold) and a judgment logic unit (comparison and judgment program). Its main function is to compare the current feature parameters with the set thresholds in real time to determine the state of the drum.

[0108] The display alarm module consists of a display unit (LCD screen, indicator lights), an audible and visual alarm device (buzzer, alarm light), and a communication output interface (Ethernet, wireless module). Its main functions are to display the strain data and operating status of each measuring point in real time, issue an audible and visual alarm when the threshold is exceeded, and then upload the alarm information to the control room or remote monitoring platform.

[0109] This disclosure provides a roller defect detection system, including a processor and a memory storing program instructions, wherein the processor is configured to execute the roller defect detection method as described in any of the above embodiments when the program instructions are executed.

[0110] The present invention has been described above by way of example, but the present invention is not limited to the specific embodiments described above. Any modifications or variations made based on the present invention shall fall within the scope of protection claimed by the present invention.

Claims

1. A method for detecting defects in a roller, characterized in that, Includes the following steps: S1, obtain the simulated strain field corresponding to different spatial positions of the roller under different working conditions, and calculate the modal coefficient of the simulated working condition based on the simulated strain field; S2, Set the set of positions for the strain sensors arranged on the roller, and calculate the local true modal coefficients corresponding to the strain sensors based on the modal coefficients of the simulation working condition; S3, the optimal modal coefficients are derived from the local true modal coefficients, and the complete strain field is reconstructed based on the optimal modal coefficients; S4. Construct a multi-objective function to determine the sensitive monitoring area and obtain the location of the strain sensor.

2. The roller defect detection method according to claim 1, characterized in that, Step S1 includes the following steps: S1.1, obtain the simulated strain field corresponding to different spatial positions of the roller under different working conditions, construct a strain snapshot matrix based on the simulated strain field, and perform mean removal processing on the strain snapshot matrix; S1.2, Perform singular value decomposition on the processed strain snapshot matrix, extract the dominant modes of the strain snapshot matrix, and calculate the low-dimensional orthogonal mode basis matrix corresponding to the dominant modes; S1.3 Calculate the modal coefficients of the simulated working condition based on the simulated strain field and the low-dimensional orthogonal modal basis matrix.

3. The drum defect detection method according to claim 2, characterized in that, Step S1.1 includes the following steps: S1.1.1 Discretize the finite element model of the roller into N spatial nodes and obtain M sets of simulated strain fields under different working conditions, including load, torque, and crack state. S1.1.2, construct the strain snapshot matrix based on the simulated strain fields of M groups. The strain snapshot matrix is: Where N is the number of spatial points and M is the number of working conditions; S1.1.3, calculate the mean of the simulated strain fields of group M, and remove the mean from the strain snapshot matrix. The calculation formula is as follows: in, , It is an identity matrix.

4. The roller defect detection method according to claim 3, characterized in that, Step S1.2 includes the following steps: S1.2.1, Perform singular value decomposition on the processed strain snapshot matrix. The decomposition formula is as follows: in, Let be a left singular matrix, representing the strain space modes; It is a diagonal matrix, and the diagonal elements are singular values, representing the magnitude of the modal energy. Let be the transpose of the right singular matrix, and let represent the transpose of the working condition coefficient matrix. This represents the energy magnitude of the k-th mode. For the k-th strain space mode, This represents the variation trend of the k-th mode under different operating conditions; S1.2.2, extract the first r dominant modes, based on the left singular matrix. Obtain the low-dimensional orthogonal mode basis matrix corresponding to the first r dominant modes. , where r is the number of truncated modes.

5. The roller defect detection method according to claim 4, characterized in that, Step S1.3 includes the following steps: S1.3.1, Set the modal coefficients for the simulation conditions. for The simulated strain field is represented based on the modal coefficients of the simulated operating conditions. Retaining the first r dominant modes, the simulated strain field is expressed as: S1.3.2, based on the simulated strain field and low-dimensional orthogonal modal basis matrix Calculate the modal coefficients of the simulation conditions. in,() + This represents the inverse operation of a matrix.

6. The roller defect detection method according to claim 5, characterized in that, Step S2 includes the following steps: S2.1, Assume p strain sensors are arranged on the roller, and the set of positions of the strain sensors is as follows: in, ; S2.2, Define the measurement point selection matrix as follows: in in, This indicates that the i-th sensor is positioned at the j-th spatial node; S2.3, Based on the measurement point selection matrix, the actual monitored strain vector is: in, Let n be the strain values ​​actually monitored by p sensors, and n be the noise, which can be ignored; combined with the low-dimensional orthogonal modal basis matrix obtained in step S1.2 The strain vector is expressed as Based on the low-dimensional orthogonal modal basis matrix Calculate the local true modal coefficients corresponding to the actual monitoring of p strain sensors. .

7. The roller defect detection method according to claim 6, characterized in that, Step S3 includes the following steps: S3.1, Given the strain vector y actually measured by the strain sensor, the local true modal coefficients are inverted using the least squares method. The optimal modal coefficients are obtained: S3.2, based on the optimal modal coefficients derived from the inversion. Reconstruct the complete strain field as 。 8. The roller defect detection method according to claim 6, characterized in that, Step S4 includes the following steps: S4.1, Based on the measurement point selection matrix in step S2.3 and the low-dimensional orthogonal modal basis matrix in step S1.2, the measurement matrix is ​​defined as follows: S4.2, Low-dimensional orthogonal modal basis matrix based on strain sensor measurement points Whether the sensitivity is balanced by setting condition number index function in, and These represent the maximum and minimum singular values ​​of the measurement matrix H; the reconstruction error function is set based on the overall deviation of the reconstructed strain field from the true strain field. in, To simulate the real strain field of the working condition, The reconstructed strain field is obtained based on the set S of p strain sensor locations; the modal coefficient error function is set based on the modal error. in, The corresponding simulation modal coefficients under the simulation conditions. , The optimal modal coefficients derived from the strain sensor Constructing a multi-objective function S4.3 Based on the multi-objective function, the sensitive monitoring area is determined by the multi-objective particle swarm optimization algorithm, and the sensor placement location is obtained.

9. The drum defect detection method according to claim 8, characterized in that, The following steps are included after step S4: S5, calculate the energy ratio, modal coefficient offset and strain gradient of the optimal modal coefficient, and determine whether the roller has cracks and locate the crack location; Step S5 includes the following steps: S5.1, calculate the proportion of the optimal modal coefficient relative to the sum of the optimal modal coefficients corresponding to the first r dominant modes, and use this as the energy proportion to determine whether a crack has occurred. The energy proportion is... in, For the k-th optimal modal coefficient, The i-th optimal modal coefficient; S5.2, Calculate the modal coefficient offset. Based on the modal coefficient offset, determine whether the roller deviates from a healthy state. The modal coefficient offset is... in, These are the modal coefficients obtained by inversion under the reference state; S5.3, Calculate the strain gradient and locate the crack position based on the strain gradient. The strain gradient is... 。 10. A roller defect detection system, characterized in that, The device includes a processor and a memory storing program instructions, characterized in that the processor is configured to execute the roller defect detection method as described in any one of claims 1 to 9 when running the program instructions.

Images