A vibration sensor layout optimization method for portable machining equipment

Abstract

The present application belongs to the technical field of mechanical structure vibration test and monitoring, and relates to a kind of portable processing equipment vibration sensor layout optimization method.The steps are as follows:S1, construct the multi-state finite element model of feeding process and public node set;S2, determine the effective modal order based on the global Fisher information matrix of multi-feeding position state fusion;S3, obtain the initial node set based on the full stroke vibration energy and radius constraint;S4, perform sequential backward screening based on the minimum information loss principle to obtain the candidate sensor node pool;S5, determine the optimal sensor arrangement scheme based on the particle swarm optimization algorithm.The present application solves the modal drift problem caused by structure time-varying by fusing multi-state characteristics, effectively avoids physical installation interference, and realizes comprehensive optimal observation of feeding full stroke vibration state.

Description

Technical Field

[0001] This invention belongs to the field of mechanical structure vibration testing and monitoring technology, and relates to a method for optimizing the layout of vibration sensors in portable processing equipment. Background Technology

[0002] With the rapid development of aerospace manufacturing technology, aircraft sub-assembly and final assembly are gradually transforming from manual operation to automation and intelligence. In this process, portable automated machining equipment, represented by automatic feed drills, is widely used in precision drilling operations for large composite materials or metal laminated components such as wings and fuselages due to its flexibility and efficiency. In actual machining, the vibration state of these portable equipment directly affects the drilling accuracy and surface quality. Therefore, constructing an efficient vibration monitoring system and acquiring high signal-to-noise ratio vibration signals through a reasonable sensor arrangement is key to achieving closed-loop control of machining quality.

[0003] However, unlike traditional large CNC machine tools or fixed mechanical structures, portable machining equipment exhibits significant "time-varying structural" characteristics. Due to their small size and light weight, the displacement of the spindle and feed mechanism during drilling and feeding causes significant changes in the overall mass distribution and stiffness matrix. This structural configuration change caused by movement may be negligible for large machine tools, but for lightweight portable equipment, it is sufficient to cause fundamental changes in its natural frequency drift and mode shapes. Currently, monitoring of such equipment lacks systematic theoretical support for sensor placement, and traditional monitoring methods based on a single static structure are ill-suited to its dynamically changing operating conditions.

[0004] Existing sensor layout optimization methods are mostly focused on linear time-invariant structures or large fixed equipment, making them difficult to directly apply to lightweight structures like portable automated machining equipment, whose configuration changes significantly with feed motion. For example, the "Method for Optimizing the Layout of Modal Analysis Measurement Points for CNC Machining Equipment" (patent number: CN102566424A) applied for by Huazhong University of Science and Technology proposes a measurement point optimization strategy based on the effective independence method for CNC machine tool structures. However, it performs modal analysis and calculation based on a single static structure, ignoring the influence of configuration changes caused by machine tool movements such as feed on modal shapes. The "A Method for Optimizing the Layout of Multimodal Vibration Sensors for Wind Tunnel Tail Support Rod" (patent number: CN120180833B) invented by Nanjing University of Technology uses a genetic algorithm to maximize modal observation capabilities and minimize the linear correlation between modal vectors. However, its research object is a cantilever beam structure with a single configuration, failing to consider the continuous evolution of dynamic characteristics caused by component displacement during operation, and thus cannot solve the problem of modal drift monitoring in portable machining equipment throughout the entire feed stroke. The "Vibration Sensor Arrangement Optimization Method Based on Complex Network Theory" (Patent No.: CN109033226B) invented by Chang'an University proposes to use complex network theory to perform mesh division and community analysis on the surface of the equipment to determine the optimal measurement points, thereby improving the comprehensive evaluation capability of the signal. However, this method relies on static vibration scanning data of the structural surface. For portable processing equipment and other dynamic equipment with significant axial displacement, its mesh division is difficult to adapt to the continuous configuration changes of the structure, and it does not consider the physical space repulsion problem when the sensor is installed.

[0005] In summary, existing sensor placement methods are limited by application scenarios (mostly static or large equipment) and theoretical assumptions (time-invariant systems), and cannot be directly applied to portable processing equipment with significant time-varying characteristics. Therefore, there is an urgent need to propose a global optimization sensor placement method that can integrate multi-state characteristics throughout the feed stroke and take into account both physical installation constraints and full-process observability. Summary of the Invention

[0006] To address the challenges of structural configuration changes, modal characteristics variations with stroke, and the inadequacy of traditional static sensor placement methods for full-process monitoring in portable machining equipment during the feeding process, this invention provides a vibration sensor layout optimization method for portable machining equipment. This method addresses the time-varying structural characteristics of portable machining equipment such as automated feed drills throughout their feeding stroke. By constructing finite element models at multiple representative feeding positions, it extracts and fuses modal shape information from different states to determine a unified and effective modal order that characterizes the dynamics throughout the entire feeding stroke. Based on this, and combining node full-stroke vibration energy, physical installation radius constraints, backward filtering of information loss, and particle swarm optimization, an optimal vibration sensor layout scheme is obtained that balances modal observability throughout the entire process, spatial installation feasibility, and modal independence.

[0007] To achieve the above objectives, the technical solution of the present invention is as follows: A method for optimizing the layout of vibration sensors in portable processing equipment includes the following steps: S1. Construct a multi-state finite element model of the feed process and a common node set; S11. Establish a multi-state finite element model; Based on the structural position change characteristics of portable machining equipment during the feed stroke, and according to the working principle, motion mode of the feed mechanism, and structural configuration change law of the portable machining equipment, representative values ​​during the feed stroke are selected. For each feed position state, a corresponding finite element model is established, where... It is a positive integer greater than 1; S12. Perform modal analysis and data extraction; right Modal analysis was performed on the finite element model at each feed position, extracting the three-dimensional spatial coordinates, modal frequencies, and mode shapes of the finite element nodes at each feed position. The mode shape data included the values ​​of the finite element nodes at each feed position. , , Mode shape components in three directions; S13. Construct a common node set; Since the finite element model of portable machining equipment may have changes in mesh node numbering under different feed positions, the finite element nodes under different feed positions are aligned based on the node numbering and the three-dimensional spatial coordinates of the finite element nodes, and a common node set shared by multiple states is established: in, Represents the set of common nodes. Indicates the number of public nodes; S2. Determine the effective mode order based on the global Fisher information matrix of multi-feed position state fusion; S21. Extract the modal shape matrix of the common node set under different feed position states; Different feed position states are denoted as , The modal truncation order is denoted as For the first In the feed position state, the first Rank mode, common node exist , , The modal components in the three directions are denoted as follows: ; in, , ; Then the first The first common node set in each feed position state The first-order mode shape matrix is ​​defined as: in, Each column represents a mode shape, and every three rows correspond to the same common node. , , Modal components in three directions; S22. Calculate the Fisher information matrix for each feed position state; Regarding the first The previous feed position state First-order mode shape matrix Calculate the corresponding Fisher information matrix: in, , used to characterize the In each feed position state, the previous The overall vibration information contained in the first mode; S23. Construct a global Fisher information matrix for multi-state fusion and calculate its norm; Introduce weight values ​​for different feed position states Construct a global Fisher information matrix: in, Indicates the first The weights of each feed position state are determined, and the following conditions are met: ; Further calculate the norm of the global Fisher information matrix: in, Denotes the 2-norm of a matrix; S24. Calculate the rate of change of the global Fisher information norm with the modal truncation order, and determine the unified effective modal order; Let the pre-defined maximum modal cutoff order be... Calculate the rate of change of the global Fisher information norm as a function of the modal truncation order: in, ; Will With modal truncation order The changes are plotted as ROC curves, and the inflection point of the ROC curve before the rapid change turns into a gradual change is selected as the effective modal order. ,in It is a positive integer greater than 1; thus ensuring that the selected modal order can cover the main dynamic characteristics under different feed position states, avoiding monitoring blind spots caused by selecting a single state order.

[0008] S3. Obtain the initial node set based on the full-stroke vibration energy and radius constraints; S31. Calculate the total vibration energy of the common node under all feed position states; Based on the modal components defined in step S21, for the common node set any public node in Calculate its in Each feed position status and forward Weighted cumulative vibrational energy under the first effective mode: in, Represents public nodes The total vibrational energy during the entire feed stroke; Different feed position state weight values ​​are defined for S23; , , They represent the first In the feed position state, the first Common nodes in the first mode exist , , Mode shape components in three directions; The unified effective modal order is determined in step S24. The larger the value, the more significant the vibration response of the node under different feed positions and main modes, and the more suitable it is as a candidate location for sensor placement.

[0009] S32. Node selection under execution radius constraint; To avoid overly dense sensor placement, spatial interference, and signal redundancy, a minimum sensor installation radius constraint is set. ; with the sum of vibrational energy Based on the selection criteria, common nodes are sorted in descending order, and the current total vibration energy is selected. The largest common node is selected as the reserved node; with this reserved node as the center, its radius is removed. Other neighboring nodes within the neighborhood, excluding the reserved node; S33, Repeating radius constraint filtering; Repeat step S32 until all common nodes have been filtered, and the final remaining common nodes constitute a set that satisfies the radius constraint. The initial set of selected nodes is denoted as: in, Represents the initial set of nodes. Indicates the number of initially selected nodes; S4. Perform backward screening of the sequence based on the principle of minimizing information loss to obtain a pool of candidate sensor nodes; S41. Construct the mode shape submatrix corresponding to the current node set; The initial set of selected nodes obtained in step S3 is used as the initial set of nodes for the backward filtering of the sequence, denoted as: in, Indicates the initial filter set; In the In this iteration, the current set of nodes is denoted as: For the For each feed position state, the modal shape matrix at that feed position state is... Extract the set of current nodes The corresponding matrix rows constitute the current node set at the th Modal mode sub-matrices for each feed position state: in, The effective modal order; Used to characterize the current node set for the first... In each feed position state, the previous The ability to observe first-order modal information; S42. Calculate the information matrix and information entropy of the current node set; Based on the modal shape sub-matrix corresponding to the current node set, construct the first... Information matrix for each feed position state: in, Used to describe the overall representational ability of the current node set for different modal information; Let information matrix eigenvalues ; Normalizing the above eigenvalues ​​yields: in, ; Further calculate the current node set at the th Information entropy at each feed position state: in, It is used to characterize the degree of balanced representation of modal information of the current node set; the more stable the information entropy, the more balanced the representation ability of the node set of different modalities.

[0010] S43. Calculate the information loss after removing a single common node; For the current set of nodes any public node in If we remove this common node, the set of nodes after the removal will be: Calculate the removal of common nodes separately Before and after, in the Information entropy change at each feed position state: in, Indicates the removal of nodes Later on the first The impact of the ability to represent individual feed position state modal information; To ensure good modal observability of the sensor layout throughout the entire feed stroke, the maximum value of the information entropy change across all feed positions is used to eliminate common nodes. The corresponding information loss: in, The smaller the value, the more likely it is that common nodes will be removed. The smaller the impact on the modal information representation capability under each feed position state, the lower the information contribution of the common node in the current node set.

[0011] S44. Perform backward elimination according to the principle of minimizing information loss; Compare the information loss values ​​of all nodes in the current node set. Select the node with the least information loss for elimination: And update the current node set: in, This indicates the node that is removed in the current iteration. Removing this node has the least impact on the ability to represent modal information throughout the feed stroke, thus reducing redundant nodes while ensuring overall observation capability.

[0012] S45. Obtain the candidate sensor node pool; Repeat steps S41 to S44 until the number of nodes in the current node set reaches the preset candidate size. The remaining nodes at this point constitute a candidate sensor node pool, denoted as: in, This represents the pool of candidate sensor nodes. Indicates the number of candidate nodes; S5. Determine the optimal sensor layout scheme based on the particle swarm optimization algorithm.

[0013] S51. Determine the optimization variables and the number of sensors; Let the final number of sensors to be deployed be... The sensor node combination to be optimized is denoted as: in, The final sensor node combination is selected from the candidate sensor node pool obtained in step S4; the number of sensors The optimal number of sensors is determined by comparing the changes in the global fitness function under different numbers of sensors. When the decrease in fitness function after increasing the number of sensors is less than the preset threshold, or when the fitness function increases in the opposite direction, the number of sensors corresponding to the inflection point of the fitness curve is selected as the optimal number of sensors. S52. Construct a multi-state modality confidence matrix; For the Each feed position state, based on the sensor node combination From the modal matrix Extract the modal vector of the corresponding node; let the first node be... The sensor nodes are combined in each state. The extracted first Rank and first The order modal vectors are as follows: Then the first In the feed position state, the first Rank and number The modal confidence value between two modes is defined as follows: in, ,and ; Further define the first The maximum value of the off-diagonal elements of the modal confidence matrix at each feed position is: in, The smaller the value, the stronger the ability of the selected sensor node combination to distinguish different modes and the lower the degree of mode aliasing.

[0014] S53. Construct the global fitness function; Introducing feed position status attention weight Construct a global fitness function that integrates multiple states: in, Indicates the first The attention weight of each feed position state in the final optimization, and satisfying: S54. Initialize the particle swarm; In the candidate sensor node pool Several particles are randomly generated in the process, each particle representing a sensor node combination scheme; the position vector of the particle represents the sensor node index combination, and the velocity vector of the particle represents the update direction and update magnitude of the node index combination. S55. Update the particle state and discretize it; Based on the velocity and position update formulas of the particle swarm optimization algorithm, the particles are iteratively updated: in, Indicates the first The particle in the first The position at the next iteration; Indicates the first The particle in the first Speed ​​during the next iteration; Indicates inertia weight; and Indicates the learning factor; and Represents a random number; Indicates the first The individual historical best position of each particle; This represents the global historical best position of the particle swarm; Since the sensor node index is a discrete variable, boundary checks, rounding mapping, and duplicate node correction are performed on the updated continuous position vector to ensure that it always corresponds to the candidate sensor node pool. The legal combination of nodes; S56, Output the optimal sensor placement position; Repeat steps S52 to S55 until the maximum number of iterations is reached or the global fitness function satisfies the preset convergence condition; extract the node combination corresponding to the globally optimal particle. It outputs the node number and three-dimensional spatial coordinates as the optimal placement position for the vibration sensor of the portable processing equipment.

[0015] The beneficial effects of this invention are: (1) The present invention does not lay out the sensor based on a single static configuration, but integrates the modal vibration information of multiple representative states throughout the feeding process. It can adapt to the changes in mass distribution, stiffness characteristics and modal vibration caused by changes in the position of the spindle, feeding mechanism or moving parts during the feeding process of portable processing equipment, thereby improving the sensor layout's adaptability to the time-varying characteristics of the structure.

[0016] (2) The present invention determines a unified effective modal order by using the global Fisher information matrix and ROC curve, so that subsequent node evaluation and sensor optimization are both established in a modal space that can cover the main dynamic characteristics of the entire journey, avoiding the monitoring blind spot caused by a single state or artificially specified modal order.

[0017] (3) The present invention introduces full-stroke vibration energy and radius constraints in the node selection stage, which can preferentially retain nodes with significant vibration response and reasonable spatial distribution, avoid overly dense sensor layout, installation interference and signal redundancy, and improve the engineering feasibility of sensor layout scheme.

[0018] (4) The present invention adopts a hierarchical optimization strategy of “energy pre-screening - information loss screening - particle swarm global optimization”, which significantly reduces the dimensionality of combinatorial optimization while ensuring the observability of multi-state modes, and improves the computational efficiency and global optimization capability of sensor layout optimization.

[0019] (5) The present invention uses the minimization of the off-diagonal elements of the multi-state modality confidence matrix as the optimization objective, which can reduce the degree of aliasing between different modes, improve the sensor network's ability to distinguish key modes, and thus improve the effectiveness and reliability of vibration monitoring signals of portable processing equipment.

[0020] (6) The sensor arrangement scheme obtained by the present invention takes into account the dynamic characteristics of the entire feed stroke, the intensity of node vibration response, spatial installation constraints and multimodal observation capabilities. It is suitable for application scenarios such as vibration monitoring, status identification, tool wear monitoring and machining quality traceability of portable processing equipment such as automatic feed drills. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the overall sensor layout optimization process of the present invention. Figure 2 The modal order diagrams based on the Fisher information matrix are shown for the three states of this invention. Figure 3 This is a fitness graph for different numbers of sensors in this invention. Detailed Implementation

[0022] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings and technical solutions.

[0023] Example 1 See Figure 1 This embodiment focuses on portable machining equipment such as automatic feed drills. Addressing the structural configuration changes caused by variations in the spindle and feed mechanism positions during the feeding process, it employs a combination of multi-state modal fusion, energy constraint screening, information loss screening, and particle swarm optimization to determine the optimal placement of vibration sensors. The specific steps are as follows.

[0024] S1. Construct a multi-state finite element model of the feed process and a common node set; Based on the feed motion characteristics of portable machining equipment, the start point, midpoint, and end point of the feed stroke are selected as three typical states, denoted as follows: , and This corresponds to the 0%, 50%, and 100% positions of the feed stroke. Finite element models for the three states were established using finite element analysis software, keeping the material parameters, connection relationships, and boundary conditions consistent across models. The structural configuration was updated only based on changes in the position of the feed mechanism.

[0025] Modal analysis was performed on the finite element model under three different conditions, and the node numbers, three-dimensional coordinates of the nodes, modal frequencies, and the values ​​of each node were extracted. , , Modal vibration components in three directions. In this embodiment, the first 20 modal data for each state are extracted in advance.

[0026] Since the finite element model may have node number offsets or local node changes under different feed position states, the finite element nodes under the three states are aligned based on the node number and the 3D coordinates of the nodes, and nodes that cannot be matched in all states are removed to construct a common node set. Subsequent modal order determination, node energy calculation, and sensor optimization are all based on this common node set.

[0027] S2. Determine the unified effective modal order; Based on the common node set, the mode shape matrices for three feed position states are extracted, and the mode shapes of each order are normalized. Then, the Fisher information matrix for each state is calculated, and state weights are introduced to construct a global Fisher information matrix. In this embodiment, all three states are equally important; therefore, the state weights are all taken as... .

[0028] To determine a unified effective modal order that covers the main dynamic characteristics of the entire feed stroke, the rate of change of the global Fisher information matrix norm with modal truncation order was calculated, and ROC curves were plotted, as shown below. Figure 2 As shown. By Figure 2 It can be seen that as the modal order increases, the ROC curve changes rapidly at first and then gradually flattens out. When the modal order increases to the 6th order, the curve variation amplitude decreases significantly, indicating that the first 6 modes can already well characterize the main dynamic features of the three feed position states. Therefore, this embodiment determines the unified effective modal order as: .

[0029] Subsequent node energy calculations, information loss screening, and particle swarm optimization are all based on the first 6 effective modes.

[0030] S3. Obtain the initial node set based on vibration energy and radius constraints; For each node in the common node set, calculate its cumulative vibration energy under three feed position states and the first six effective modes. The larger the cumulative vibration energy of a node, the more significant its vibration response is throughout the entire feed stroke, and the more suitable it is as a candidate location for sensor placement.

[0031] To avoid overly dense sensor placement, spatial interference, and signal redundancy, a minimum sensor installation radius constraint is set. Nodes are sorted from largest to smallest based on their accumulated vibration energy, with priority given to retaining the node with the highest vibration energy and discarding those with the largest radius. Other nodes within the neighborhood. Repeat the above process until all nodes have been filtered, resulting in a preliminary set of nodes that satisfy the spatial distance constraints. .

[0032] S4. Obtain a pool of candidate sensor nodes based on the principle of minimizing information loss; In the initial node set Based on this, a sequential backward selection algorithm is used to further filter nodes. In each iteration, the impact of removing a certain node on the modal information representation capability under the three feed position states is examined in turn, and the change in information entropy before and after removal is used as the information loss evaluation index.

[0033] For any node to be removed, calculate its... , and The change in information entropy under each state is used, and the largest change among the three states is taken as the removal loss for that node. The smaller the removal loss, the smaller the contribution of that node to the multi-state modal information. The node with the smallest removal loss is removed in each round, and this process is repeated until the number of remaining nodes reaches a preset candidate size, thus obtaining a candidate sensor node pool. .

[0034] This step reduces the search dimensionality of subsequent particle swarm optimization while ensuring the observability of the feed mode throughout its entire stroke, thus improving optimization efficiency.

[0035] S5. Determine the optimal sensor layout scheme based on particle swarm optimization; In the candidate sensor node pool In this study, a particle swarm optimization algorithm is used to search for the final sensor node combination. Each particle represents a sensor node combination scheme, and the particle's position vector represents the sensor node index combination.

[0036] In this embodiment, the maximum value of the off-diagonal elements of the multi-state modality confidence matrix is ​​used as the optimization evaluation index. For any sensor node combination, its MAC matrix is ​​calculated in three feed position states, and the maximum value of the off-diagonal elements of the MAC matrix in each state is extracted. The smaller this value, the stronger the sensor node combination's ability to distinguish different modes and the lower the degree of mode aliasing.

[0037] Furthermore, equal weights are introduced for the three states to construct a global fitness function. During particle swarm optimization, the individual optimal solution and the global optimal solution of each particle are continuously updated, and boundary checks, rounding mappings, and duplicate node corrections are performed on the updated node indices to ensure that each particle always corresponds to a legal combination of nodes in the candidate node pool.

[0038] This embodiment performs optimization calculations for cases where the number of sensors increases from 2 to 20, and plots fitness change curves for different numbers of sensors, such as... Figure 3 As shown. By Figure 3It can be seen that when the number of sensors is small, the fitness value is high, indicating a greater risk of mode aliasing. As the number of sensors increases, the fitness value gradually decreases, and the mode discrimination capability of the sensor network is enhanced. When the number of sensors continues to increase, the rate of decrease in fitness decreases, and it may even stop improving significantly due to the introduction of correlation by redundant nodes.

[0039] Taking into account modal discrimination capability, installation complexity, number of sensors, and engineering feasibility, this embodiment ultimately determines the number of sensors to be 8. After the particle swarm optimization algorithm converges, the node numbers corresponding to the 8 globally optimal particles are: The three-dimensional spatial coordinates corresponding to the above node numbers are the final installation positions of the vibration sensor.

[0040] Verification shows that the maximum off-diagonal element of the MAC matrix of this sensor group is less than 0.25 in all three states before, during and after feeding, proving that the method can effectively overcome the modal drift problem caused by time-varying structures and achieve high-precision vibration monitoring throughout the entire stroke.

Claims

1. A method for optimizing the layout of vibration sensors in portable processing equipment, characterized in that, Includes the following steps: S1. Construct a multi-state finite element model of the feed process and a common node set; S2. Determine the effective mode order based on the global Fisher information matrix of multi-feed position state fusion; S3. Obtain the initial node set based on the full-stroke vibration energy and radius constraints; S4. Perform backward screening of the sequence based on the principle of minimizing information loss to obtain a pool of candidate sensor nodes; S5. Determine the optimal sensor layout scheme based on the particle swarm optimization algorithm.

2. The method for optimizing the layout of vibration sensors in portable processing equipment according to claim 1, characterized in that, The specific implementation process of step S1 is as follows: S11. Establish a multi-state finite element model; Based on the structural position change characteristics of portable machining equipment during the feed stroke, and according to the working principle, motion mode of the feed mechanism, and structural configuration change law of the portable machining equipment, representative values ​​during the feed stroke are selected. For each feed position state, a corresponding finite element model is established, where... It is a positive integer greater than 1; S12. Perform modal analysis and data extraction; right Modal analysis was performed on the finite element model at each feed position, extracting the three-dimensional spatial coordinates, modal frequencies, and mode shapes of the finite element nodes at each feed position. The mode shape data included the values ​​of the finite element nodes at each feed position. , , Mode shape components in three directions; S13. Construct a common node set; Since the finite element model of portable machining equipment may have changes in mesh node numbering under different feed positions, the finite element nodes under different feed positions are aligned based on the node numbering and the three-dimensional spatial coordinates of the finite element nodes, and a common node set shared by multiple states is established: in, Represents the set of common nodes. This indicates the number of public nodes.

3. The method for optimizing the layout of vibration sensors in portable processing equipment according to claim 2, characterized in that, The specific implementation process of step S2 is as follows: S21. Extract the modal shape matrix of the common node set under different feed position states; Different feed position states are denoted as , The modal truncation order is denoted as For the first In the feed position state, the first Rank mode, common node exist , , The modal components in the three directions are denoted as follows: ; in, , ; Then the first The first common node set in each feed position state The first-order mode shape matrix is ​​defined as: in, Each column represents a mode shape, and every three rows correspond to the same common node. , , Modal components in three directions; S22. Calculate the Fisher information matrix for each feed position state; Regarding the first The previous feed position state First-order mode shape matrix Calculate the corresponding Fisher information matrix: in, , used to characterize the In each feed position state, the previous The overall vibration information contained in the first mode; S23. Construct a global Fisher information matrix for multi-state fusion and calculate its norm; Introduce weight values ​​for different feed position states Construct a global Fisher information matrix: in, Indicates the first The weights of each feed position state are determined, and the following conditions are met: ; Further calculate the norm of the global Fisher information matrix: in, Denotes the 2-norm of a matrix; S24. Calculate the rate of change of the global Fisher information norm with the modal truncation order, and determine the unified effective modal order; Let the pre-defined maximum modal cutoff order be... Calculate the rate of change of the global Fisher information norm as a function of the modal truncation order: in, ; Will With modal truncation order The changes are plotted as ROC curves, and the inflection point of the ROC curve before the rapid change turns into a gradual change is selected as the effective modal order. ,in It is a positive integer greater than 1.

4. The method for optimizing the layout of vibration sensors in portable processing equipment according to claim 3, characterized in that, The specific implementation process of step S3 is as follows: S31. Calculate the total vibration energy of the common node under all feed position states; Based on the modal components defined in step S21, for the common node set any public node in Calculate its in Each feed position status and forward Weighted cumulative vibrational energy under the first effective mode: in, Represents public nodes The total vibrational energy during the entire feed stroke; S32. Node selection under execution radius constraint; To avoid overly dense sensor placement, spatial interference, and signal redundancy, a minimum sensor installation radius constraint is set. ; with the sum of vibrational energy Based on the selection criteria, common nodes are sorted in descending order, and the current total vibration energy is selected. The largest common node is selected as the reserved node; with this reserved node as the center, its radius is removed. Other neighboring nodes within the neighborhood, excluding the reserved node; S33, Repeating radius constraint filtering; Repeat step S32 until all common nodes have been filtered, and the final remaining common nodes constitute a set that satisfies the radius constraint. The initial set of selected nodes is denoted as: in, Represents the initial set of nodes. This indicates the number of nodes initially selected.

5. The method for optimizing the layout of vibration sensors in portable processing equipment according to claim 4, characterized in that, The specific implementation process of step S4 is as follows: S41. Construct the mode shape submatrix corresponding to the current node set; The initial set of selected nodes obtained in step S3 is used as the initial set of nodes for the backward filtering of the sequence, denoted as: in, Indicates the initial filter set; In the In this iteration, the current set of nodes is denoted as: For the For each feed position state, the modal shape matrix at that feed position state is... Extract the set of current nodes The corresponding matrix rows constitute the current node set at the th Modal mode sub-matrices for each feed position state: in, The effective modal order; Used to characterize the current node set for the first... In each feed position state, the previous The ability to observe first-order modal information; S42. Calculate the information matrix and information entropy of the current node set; Based on the modal shape sub-matrix corresponding to the current node set, construct the first... Information matrix for each feed position state: in, Used to describe the overall representational ability of the current node set for different modal information; Let information matrix eigenvalues ; Normalizing the above eigenvalues ​​yields: in, ; Further calculate the current node set at the th Information entropy at each feed position state: in, Used to characterize the degree of balanced representation of modal information of the current node set; S43. Calculate the information loss after removing a single common node; For the current set of nodes any public node in If we remove this common node, the set of nodes after the removal will be: Calculate the removal of common nodes separately Before and after, in the Information entropy change at each feed position state: in, Indicates the removal of nodes Later on the first The impact of the ability to represent individual feed position state modal information; Take the maximum value of the information entropy change across all feed position states as the common node to be removed. The corresponding information loss: S44. Perform backward elimination according to the principle of minimizing information loss; Compare the information loss values ​​of all nodes in the current node set. Select the node with the least information loss for elimination: And update the current node set: in, This indicates the node that was removed in the current iteration; S45. Obtain the candidate sensor node pool; Repeat steps S41 to S44 until the number of nodes in the current node set reaches the preset candidate size. The remaining nodes at this point constitute a candidate sensor node pool, denoted as: in, This represents the pool of candidate sensor nodes. This indicates the number of candidate nodes.

6. The method for optimizing the layout of vibration sensors in portable processing equipment according to claim 5, characterized in that, The specific implementation process of step S5 is as follows: S51. Determine the optimization variables and the number of sensors; Let the final number of sensors to be deployed be... The sensor node combination to be optimized is denoted as: in, The final sensor node combination is selected from the candidate sensor node pool obtained in step S4; the number of sensors The optimal number of sensors is determined by comparing the changes in the global fitness function under different numbers of sensors. When the decrease in fitness function after increasing the number of sensors is less than the preset threshold, or when the fitness function increases in the opposite direction, the number of sensors corresponding to the inflection point of the fitness curve is selected as the optimal number of sensors. S52. Construct a multi-state modality confidence matrix; For the Each feed position state, based on the sensor node combination From the modal matrix Extract the modal vector of the corresponding node; let the first node be... The sensor nodes are combined in each state. The extracted first Rank and first The order modal vectors are as follows: Then the first In the feed position state, the first Rank and number The modal confidence value between two modes is defined as follows: in, ,and ; Further define the first The maximum value of the off-diagonal elements of the modal confidence matrix at each feed position is: S53. Construct the global fitness function; Introducing feed position status attention weight Construct a global fitness function that integrates multiple states: in, Indicates the first The attention weight of each feed position state in the final optimization, and satisfying: S54. Initialize the particle swarm; In the candidate sensor node pool Several particles are randomly generated in the process, each particle representing a sensor node combination scheme; the position vector of the particle represents the sensor node index combination, and the velocity vector of the particle represents the update direction and update magnitude of the node index combination. S55. Update the particle state and discretize it; Based on the velocity and position update formulas of the particle swarm optimization algorithm, the particles are iteratively updated: in, Indicates the first The particle in the first The position at the next iteration; Indicates the first The particle in the first Speed ​​during the next iteration; Indicates inertia weight; and Indicates the learning factor; and Represents a random number; Indicates the first The individual historical best position of each particle; This represents the global historical best position of the particle swarm; Since the sensor node index is a discrete variable, boundary checks, rounding mapping, and duplicate node correction are performed on the updated continuous position vector to ensure that it always corresponds to the candidate sensor node pool. The legal combination of nodes; S56, Output the optimal sensor placement position; Repeat steps S52 to S55 until the maximum number of iterations is reached or the global fitness function satisfies the preset convergence condition; extract the node combination corresponding to the globally optimal particle. It outputs the node number and three-dimensional spatial coordinates as the optimal placement position for the vibration sensor of the portable processing equipment.

Images