PCB eddy current detection coil optimization method and system based on maximum defect response amplitude

Abstract

The application provides a PCB eddy current detection coil optimization method and system based on maximum defect response amplitude, belongs to the technical field of PCB eddy current detection coil optimization, and comprises the following steps: a three-dimensional electromagnetic finite element parameterized simulation model is established; a defect response amplitude evaluation index is defined to represent the sensitivity of a probe to defects by comparing received signals in healthy and defect states; a frequency-layer number two-dimensional parameter scanning space is constructed and automatically batch simulated; in the process of automatic batch simulation, defect response characteristics of each parameter combination in the frequency-layer number two-dimensional parameter scanning space are calculated to form a two-dimensional defect response matrix; a multilayer feedforward neural network proxy model is constructed with frequency and layer number as inputs and defect response amplitude as predicted value output; and the predicted value of the multilayer feedforward neural network proxy model is used as a fitness function, and a particle swarm optimization algorithm is used to search for an optimal excitation frequency and an optimal PCB coil layer number capable of generating the maximum defect response amplitude.

Description

Technical Field

[0001] This invention belongs to the field of PCB eddy current detection coil optimization technology, and particularly relates to a PCB eddy current detection coil optimization method and system based on maximizing defect response amplitude. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] In critical sectors such as petrochemicals, nuclear power, and urban water supply, metal pipelines serve as core infrastructure for fluid transmission, bearing immense operational pressure. However, due to prolonged exposure to complex environments involving high pressure, corrosive media, or alternating loads, pipelines are highly susceptible to damage such as wall thinning, fatigue cracking, or pitting corrosion. Regularly conducting high-precision non-destructive testing to provide early warning of potential risks is essential for ensuring the safe operation of industrial systems throughout their entire lifecycle.

[0004] Among numerous non-destructive testing technologies, eddy current testing, utilizing the principle of electromagnetic induction, offers significant advantages such as non-contact operation, no pretreatment required, high testing speed, and high sensitivity to surface and near-surface defects. In automated online pipeline inspection, eddy current technology can penetrate non-conductive coatings to capture minute crack signals, making it the preferred solution for rapid integrity assessment of metal pipelines.

[0005] As sensors become increasingly miniaturized and flexible, eddy current coils on printed circuit boards (PCBs) are gradually replacing traditional wire-wound coils due to their high geometric consistency, low production cost, and ease of fitting to irregularly shaped surfaces. However, the performance of PCB probes based on eddy current coils is constrained by both their physical parameters and excitation conditions. While increasing the number of coil layers can enhance the induced magnetic field, it also introduces parasitic capacitance and alters the coil impedance. Furthermore, adjusting the excitation frequency directly affects the skin depth of the eddy current and the intensity of defect disturbances. This complex nonlinear coupling between frequency and layer number makes it difficult to explore the probe's ultimate detection capabilities using traditional optimization methods that rely on experience or single variables. Summary of the Invention

[0006] To overcome the shortcomings of the prior art, this invention provides a method and system for optimizing PCB eddy current detection coils based on maximizing defect response amplitude. By utilizing the ANSYS electromagnetic simulation model and the multilayer feedforward neural network surrogate model, and based on the particle swarm optimization algorithm (PSO), a global search is performed in the continuous parameter space to achieve the optimal parameter combination configuration of the excitation frequency and the number of coil layers of the PCB eddy current detection probe, thereby improving the probe's detection sensitivity for metal defects.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: Firstly, a method for optimizing PCB eddy current detection coils based on maximizing defect response amplitude is disclosed, including: A three-dimensional electromagnetic finite element parametric simulation model was established, with the excitation frequency and the number of PCB coil layers as variables to be optimized. Based on a three-dimensional electromagnetic finite element parametric simulation model, by comparing the received signals under healthy and defective states, a defect response amplitude evaluation index is defined to characterize the probe's sensitivity to defects. The excitation frequency and the number of PCB coil layers are discretized to construct a two-dimensional parameter scanning space of frequency-layer number and perform automatic batch simulation. During the automated batch simulation process, the defect response characteristics under various parameter combinations in the frequency-layer two-dimensional parameter scanning space are calculated and a two-dimensional defect response matrix is ​​formed. A multi-layer feedforward neural network surrogate model is constructed with frequency and layer number as inputs and defect response amplitude as predicted output. A nonlinear mapping relationship is established, and the model is trained based on the two-dimensional defect response matrix to obtain the trained multi-layer feedforward neural network surrogate model. Using the prediction values ​​of the multilayer feedforward neural network surrogate model as the fitness function, the particle swarm optimization algorithm is used to search for the optimal excitation frequency and the optimal number of PCB coil layers that can produce the maximum defect response amplitude.

[0008] As a further technical solution, when establishing a three-dimensional electromagnetic finite element parametric simulation model, the PCB coil is parametrically modeled, and simulation models under different frequencies and different number of layers are automatically generated. During the simulation process, an adaptive mesh generation method is used to locally refine the coil area, defect area, and near-field electromagnetic coupling area.

[0009] As a further technical solution, when defining the defect response amplitude evaluation index, the defect disturbance signal is defined as the difference between the received signal in the defect state and the received signal in the healthy state of the three-dimensional electromagnetic finite element parameterized simulation model, and the maximum value of the defect disturbance signal amplitude is defined as the defect response amplitude evaluation index.

[0010] As a further technical solution, the process of forming a two-dimensional defect response matrix is ​​as follows: For each set of parameters, apply the corresponding frequency excitation signal and calculate the receive response under healthy conditions; then calculate the receive response under defective conditions. Calculate the defect disturbance signal; Extract the peak value of the defect response from the defect disturbance signal; After calculating all parameter combinations, a two-dimensional defect response matrix is ​​formed, where the rows correspond to different excitation frequencies and the columns correspond to different PCB layer numbers; the matrix elements represent the defect response amplitude under the corresponding parameter combination.

[0011] As a further technical solution, it also includes: training and validating a multi-layer feedforward neural network surrogate model so that it can accurately predict defect response values ​​in a continuous space. Specifically, during training: The two-dimensional defect response matrix data is divided into training set, validation set and test set, and mean squared error is used as the loss function. The Adam optimization algorithm is used to update network parameters during training. Once the validation set error meets the convergence condition, the neural network training is complete.

[0012] As a further technical solution, when using the particle swarm optimization algorithm to search for the optimal excitation frequency and the optimal number of PCB coil layers that can produce the maximum defect response amplitude, the following are included: Using a trained neural network model, the magnitude of the defect response is predicted in a continuous parameter space; An optimization objective is established based on the optimal excitation frequency and the optimal number of PCB coil layers; The particle swarm optimization algorithm is used to perform a global optimization search on the frequency-layer parameter space: First, the position of each particle will be defined; The particle's velocity vector is further defined as follows: Subsequently, the defect response amplitude predicted by the neural network was used as the particle fitness function; After initializing the particle swarm, multiple particle positions are randomly generated within a preset frequency range and layer number range, and the following are recorded: the historical best position and the global best position for each particle. In each iteration, the particle velocity is updated based on the individual particle's optimal position and the global optimal position; Then update the particle position; The updated particle positions are then input into the neural network model and the particle fitness is recalculated. When a particle's current fitness is better than its historical best fitness, update the individual's best position; when a particle's current fitness is better than the global best fitness, update the global best position. Ultimately, the globally optimal particle position is taken as the optimization result: that is, the excitation frequency that can produce the maximum defect response amplitude is selected in combination with the PCB coil layer array as the final optimization parameters.

[0013] Secondly, a PCB eddy current detection coil optimization system based on maximizing defect response amplitude is disclosed, including: The simulation model building module is configured to: establish a three-dimensional electromagnetic finite element parametric simulation model, with the excitation frequency and the number of PCB coil layers as variables to be optimized; The defect response amplitude evaluation index definition module is configured to: define a defect response amplitude evaluation index to characterize the probe’s sensitivity to defects by comparing the received signals in healthy and defective states, based on a three-dimensional electromagnetic finite element parametric simulation model. The batch simulation module is configured to: discretize the excitation frequency and the number of PCB coil layers, construct a two-dimensional parameter scanning space of frequency-layer number, and automatically perform batch simulation; The two-dimensional defect response matrix forming module is configured to: calculate the defect response characteristics under various parameter combinations in the frequency-layer two-dimensional parameter scan space and form a two-dimensional defect response matrix during the automatic batch simulation process. The surrogate model construction module is configured to: construct a multi-layer feedforward neural network surrogate model with frequency and layer number as inputs and defect response amplitude as predicted output; establish a nonlinear mapping relationship; train the model based on the two-dimensional defect response matrix; and obtain the trained multi-layer feedforward neural network surrogate model. The module for determining the optimal excitation frequency and the optimal number of PCB coil layers is configured to use the predicted value of the multilayer feedforward neural network surrogate model as the fitness function and the particle swarm optimization algorithm to search for the optimal excitation frequency and the optimal number of PCB coil layers that can produce the maximum defect response amplitude.

[0014] The above one or more technical solutions have the following beneficial effects: The proposed method for joint optimization of PCB eddy current detection coil frequency and layer number based on maximizing defect response amplitude breaks through the limitations of traditional design relying on single-variable tuning. By deeply integrating finite element electromagnetic simulation, neural network nonlinear modeling, and heuristic swarm intelligence search algorithm, this method constructs a closed-loop design framework that starts from physical mechanisms, is accelerated by data, and is ultimately decided by optimization algorithms, achieving precise and coordinated configuration of the core parameters of the PCB probe in a continuous solution space.

[0015] This invention directly optimizes for maximizing the defect response amplitude, locking in a balance between the excitation frequency and the number of coil layers. Through synergistic optimization, it maximizes the disturbance effect of defects on the electromagnetic field, resulting in higher signal resolution and signal-to-noise ratio when dealing with micro-cracks or deep defects, ensuring the accuracy of the detection results.

[0016] This invention utilizes a neural network surrogate model to replace the traditional time-consuming full-space finite element iterative calculation. After acquiring a small amount of sample data, the neural network can instantly predict the detection performance under any combination of parameters, reducing the simulation search process that originally required several days or even weeks to the second level, greatly improving R&D efficiency and reducing the consumption of software and hardware resources.

[0017] Compared to the traditional discrete parameter scanning method, this invention introduces a particle swarm optimization algorithm to perform global optimization in a continuous frequency and layer space, providing a more scientific and robust design solution for the development of high-performance, customized PCB eddy current sensors.

[0018] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0019] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0020] Figure 1 This is a flowchart of the PCB eddy current detection coil frequency-layer joint optimization method based on maximizing defect response amplitude according to an embodiment of the present invention; Figure 2 Schematic diagrams of PCB coils with different numbers of layers; Figure 3 Schematic diagram of electromagnetic response signals under different states; Figure 4 This is a schematic diagram of a two-dimensional defect response; Figure 5 A schematic diagram of the constructed neural network model; Figure 6 This is a schematic diagram of the particle swarm optimization algorithm. Detailed Implementation

[0021] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0022] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0023] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0024] For optimization problems involving complex electromagnetic models, artificial intelligence (AI) technology offers an efficient solution. Neural networks possess exceptional nonlinear mapping capabilities, enabling them to build surrogate models between parameters and responses by learning from limited simulation data, thus avoiding the extremely high time costs of large-scale finite element method (FEM) calculations. Meanwhile, particle swarm optimization (PSO), as a mature heuristic global search technique, features fast convergence and robustness, enabling it to quickly locate the global optimum within the continuous response space constructed by the neural network. The combination of these two approaches provides a complete logical framework for solving multivariate collaborative optimization problems.

[0025] This invention proposes a joint optimization method for the frequency and number of layers of PCB eddy current detection coils based on maximizing the defect response amplitude. First, feature data is obtained by establishing an accurate finite element simulation model. Then, a high-precision performance prediction surrogate model is constructed using a multilayer feedforward neural network. Finally, a particle swarm optimization algorithm is introduced to perform intelligent search across the entire parameter space. This integrated approach yields the probe parameter combination with the strongest defect detection capability, providing a scientific design basis for the development of high-performance PCB eddy current sensors.

[0026] Example 1 This embodiment discloses a PCB eddy current detection coil optimization method based on maximizing defect response amplitude. The detection process of this method is as follows: Figure 1 As shown, it includes the following steps: Step (1): Build a PCB eddy current testing system and use ANSYS to create a three-dimensional electromagnetic finite element parametric simulation model that includes the excitation coil, the test piece and the defect area; Step (2): Based on the three-dimensional electromagnetic finite element parameterized simulation model, by comparing the received signals in the healthy state and the defective state, a defect response amplitude evaluation index is defined to characterize the sensitivity of the probe to defects. Step (3): Discretize the excitation frequency and the number of coil layers, construct a two-dimensional parameter scanning space, and write a script to realize automatic batch simulation. Here, the excitation frequency and the number of coil layers are preset ranges obtained through experience. Step (4): Conduct large-scale parameter scanning simulation, calculate the defect response characteristics under each parameter combination, and form a two-dimensional defect response matrix; Step (5): Construct a multi-layer feedforward neural network surrogate model with frequency and layer number as input and defect response amplitude as output, and establish a nonlinear mapping relationship; Step (6): The Adam optimization algorithm is used to train and validate the neural network so that it can accurately predict the defect response value in the continuous space; Step (7): Using the neural network prediction value as the fitness function, the particle swarm optimization algorithm is used to search for the optimal frequency and optimal number of layers that can produce the maximum defect response amplitude; Step (8): Output the optimal parameter combination result and design and process the PCB eddy current detection probe based on it to complete the optimized detection for metal structures.

[0027] In this implementation example, an ANSYS electromagnetic simulation model is established, evaluation indicators are confirmed, and parameters are optimized based on the evaluation indicators. Specifically, the defect response patterns under different excitation frequencies and different PCB coil layer numbers are analyzed, and the optimal parameter search is achieved by combining a neural network surrogate model, thereby obtaining the PCB eddy current detection probe parameter combination with the maximum defect response capability.

[0028] In one implementation example, in step (1): a PCB eddy current detection system is built and an ANSYS simulation model is established. Parameter optimization is performed in the simulation, and the conclusions of the final simulation are used to guide the structural design of the system.

[0029] First, a PCB eddy current testing system was built, and a three-dimensional electromagnetic finite element simulation model was created using ANSYS. PCB coils with different layers were then compared. Figure 2 As shown, where Figure 2 (a) in the diagram represents a PCB coil with two layers. Figure 2 (b) in the diagram represents a PCB coil with 4 layers. Figure 2 (c) in the diagram represents an 8-layer PCB coil. The PCB eddy current testing system mainly includes: a PCB planar spiral excitation coil, a metal test piece, a defect area, an excitation current source, a signal receiving module, and air regions and boundary conditions. The PCB planar spiral excitation coil is placed above the test piece, with the defect located within the test piece at the corresponding position below the coil. The excitation current flows through the coil, and the receiving module detects the output signal of the coil. Defect area: a 5mm*5mm square defect with a depth of 2mm; Air region: the solid structure extending outwards by 200% is defined as air; Boundary conditions: physical constraints on the boundary of the region, used to simulate an infinitely large space.

[0030] In this implementation example, the PCB coil adopts a planar spiral structure, and various coil models are constructed by setting different PCB layer counts. The test piece is made of conductive metal material, and an artificial defect structure is preset inside the test piece. The defect area is a 5mm*5mm square defect with a depth of 2mm. Among them, the excitation frequency and the number of PCB coil layers are variables to be optimized.

[0031] Let the excitation frequency be... The number of PCB coil layers is Therefore, the frequency search range selected based on experience is:

[0032] The PCB layer search range is:

[0033] Subsequently, parametric modeling of the PCB coil was performed, and simulation models under different frequencies and layer numbers were automatically generated in ANSYS. During the simulation, an adaptive mesh generation method was used to locally refine the coil region, defect region, and near-field electromagnetic coupling region. The adaptive mesh generation method was used to ensure the calculation accuracy in key areas with drastic changes in field strength gradient while avoiding the waste of computational resources caused by blindly refining the entire field, thereby improving the calculation accuracy of high-frequency eddy current distribution.

[0034] In one implementation example, in step (2): a defect response amplitude evaluation index is established, which is used to characterize the maximum disturbance caused by the defect to the electromagnetic field distribution. When the defect response amplitude is larger, it indicates that the probe is more sensitive to the defect and has a stronger detection capability.

[0035] Based on the simulation model in step 1, electromagnetic response models for healthy and defective states are established respectively.

[0036] Let the received signal in the coil's healthy state be: The received signal under defective conditions is: f is a function of frequency, N is the number of layers, and t is time. Signals in different states are as follows: Figure 3 As shown, the defect disturbance signal is defined as follows:

[0037] The defect response amplitude evaluation index is further defined as follows:

[0038] in, This represents the peak value of the defect response under the corresponding frequency and layer number conditions; Indicates the sampling time; This indicates an absolute value operation. This index is used to characterize the maximum disturbance that a defect causes to the electromagnetic field distribution. The larger the defect response amplitude, the more sensitive the probe is to the defect and the stronger its detection capability.

[0039] In one implementation example, in step (3): construct the frequency-layer two-dimensional parameter scan space, i.e., construct the matrix.

[0040] The excitation frequency and the number of PCB coil layers are discretized. The frequency sampling set is represented as:

[0041] The layer sampling set is represented as:

[0042] in, The number of frequency sampling points. This represents the number of sampling points for the PCB layers.

[0043] Then, a two-dimensional parameter combination is constructed:

[0044] Finally, batch finite element simulation analysis is performed by automatically calling ANSYS-based simulation models using parametric scripts.

[0045] In one implementation example, in step (4): a parameter scan is performed and a defect response matrix is ​​formed.

[0046] For each set of parameter combinations The following processes will be carried out in sequence: Apply the corresponding frequency excitation signal to the coil and calculate the receive response under healthy conditions:

[0047] The received response under defective condition was then obtained:

[0048] Next, the defect disturbance signal is calculated:

[0049] Extracting the peak value of the defect response:

[0050] After completing the calculation for all parameter combinations, the result is as follows: Figure 4 The two-dimensional defect response matrix shown:

[0051] In this matrix, the row direction corresponds to different excitation frequencies, and the column direction corresponds to different PCB layer numbers; the elements of the two-dimensional defect response matrix represent the defect response amplitude under the corresponding parameter combination.

[0052] In this implementation example, in step (5): a neural network agent model is constructed.

[0053] To reduce the computational complexity caused by large-scale finite element parameter scanning, the following approach is adopted: Figure 5 The neural network model shown establishes a nonlinear mapping relationship between frequency, number of layers, and defect response.

[0054] The input parameters of a neural network are defined as follows:

[0055] Network output is defined as:

[0056] The input layer contains two neurons, corresponding to the excitation frequency and the number of PCB layers, respectively; the output layer contains one neuron, corresponding to the defect response amplitude prediction result.

[0057] The neural network employs a multi-layer feedforward neural network structure, including: The system consists of an input layer, a first hidden layer, a second hidden layer, and an output layer. The first hidden layer contains 64 neurons, and the second hidden layer contains 32 neurons. The ReLU activation function is used for the hidden layers.

[0058] The forward propagation process of the network is represented as: Output of the first hidden layer:

[0059] Second hidden layer output:

[0060] The final output layer result is:

[0061] in, The weight matrix is ​​obtained from training the neural network; This is the bias vector obtained from training the neural network. To predict the magnitude of the defect response.

[0062] In one implementation example, in step (6): neural network training.

[0063] The two-dimensional defect response matrix data is divided into training, validation, and test sets. Mean squared error is used as the loss function.

[0064] in, This represents the number of training samples; This represents the actual defect response value; This is the network prediction value. This represents the set of network parameters.

[0065] The Adam optimization algorithm is used to update network parameters during training.

[0066] The network parameter update process is represented as follows:

[0067] in, The learning rate; This is the gradient of the loss function. t、 t+1 represents the network parameters at times t and t+1, respectively.

[0068] Once the validation set error meets the convergence condition, the neural network training is complete.

[0069] In one implementation example, in step (7): the optimal frequency and optimal layer number are searched based on particle swarm optimization.

[0070] Using a trained neural network model, the magnitude of the defect response is predicted in a continuous parameter space that includes combinations of different frequencies and layers.

[0071] in This represents the completed training of the neural network model. Indicates the excitation frequency; Indicates the number of coil layers on the PCB; This indicates the predicted defect response amplitude.

[0072] Further establish optimization goals:

[0073] in, To achieve the optimal excitation frequency; The optimal number of PCB coil layers.

[0074] To improve the efficiency of parameter search and avoid the increased computational cost associated with traditional traversal search, a method such as... Figure 6 The particle swarm optimization algorithm shown performs a global optimization search on the frequency-layer parameter space.

[0075] First, the position of each particle is defined as:

[0076] in: Indicates the first The parameter positions of each particle, Indicates the corresponding excitation frequency; This indicates the number of coil layers on the corresponding PCB.

[0077] The particle's velocity vector is further defined as:

[0078] in: Indicates the frequency search speed; This indicates the search speed based on the number of layers.

[0079] Subsequently, the defect response amplitude predicted by the neural network was used as the particle fitness function:

[0080] After initializing the particle swarm, multiple particle positions are randomly generated within a preset frequency range and layer number range, and recorded: The historical best position of each particle Global optimal position .

[0081] In each iteration, the particle velocity is updated based on the individual particle's optimal position and the global optimal position:

[0082] in: Inertial weight; For individual learning factors; For group learning factors; It is a random number; This indicates the current iteration number.

[0083] Then update the particle position:

[0084] The updated particle positions are then input into the neural network model. And recalculate particle fitness.

[0085] When a particle's current fitness is better than its historical best fitness, update the individual's optimal position. Update the global optimal position when the particle's current fitness is better than the global optimal fitness. .

[0086] Repeat the above iterative process until one of the following conditions is met: (1) Reach the maximum number of iterations; (2) The change in global fitness is less than the preset threshold; (3) Particle swarm convergence.

[0087] Finally, the globally optimal particle position is taken as the optimization result: That is, the excitation frequency that can produce the maximum defect response amplitude is selected in combination with the PCB coil layer array as the final optimization parameters.

[0088] In one implementation example, in step (8): the optimization results are output and the probe design is completed.

[0089] Based on the optimal excitation frequency obtained through optimization With the optimal number of PCB layers The design of the PCB eddy current detection probe was completed; based on the optimization results, the PCB coil was processed and applied to the detection of metal structure defects. The probe design and PCB coil processing are existing technologies and will not be described in detail here.

[0090] In this implementation example, Figure 1 , Figure 6 This paper describes the complete detection process, core algorithm logic, and parameter design criteria of a PCB eddy current testing coil frequency-layer joint optimization method based on maximizing defect response amplitude. It utilizes a neural network to construct a nonlinear surrogate model relating excitation frequency, PCB coil layer number, and defect response amplitude, enabling a mapping method from discrete simulation samples to continuous parameter space performance prediction. A particle swarm optimization algorithm is employed for global optimization within the neural network prediction space, using maximizing defect response amplitude as the objective function to automatically retrieve the optimal excitation frequency and optimal coil layer number.

[0091] Example 2 The purpose of this embodiment is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described method.

[0092] Example 3 The purpose of this embodiment is to provide a computer-readable storage medium.

[0093] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the above method.

[0094] Example 4 The purpose of this embodiment is to provide a PCB eddy current detection coil optimization system based on maximizing defect response amplitude, including: The simulation model building module is configured to: establish a three-dimensional electromagnetic finite element parametric simulation model, with the excitation frequency and the number of PCB coil layers as variables to be optimized; The defect response amplitude evaluation index definition module is configured to: define a defect response amplitude evaluation index to characterize the probe’s sensitivity to defects by comparing the received signals in healthy and defective states, based on a three-dimensional electromagnetic finite element parametric simulation model. The batch simulation module is configured to: discretize the excitation frequency and the number of PCB coil layers, construct a two-dimensional parameter scanning space of frequency-layer number, and automatically perform batch simulation; The two-dimensional defect response matrix forming module is configured to: calculate the defect response characteristics under various parameter combinations in the frequency-layer two-dimensional parameter scan space and form a two-dimensional defect response matrix during the automatic batch simulation process. The surrogate model construction module is configured to: construct a multi-layer feedforward neural network surrogate model with frequency and layer number as inputs and defect response amplitude as predicted output; establish a nonlinear mapping relationship; train the model based on the two-dimensional defect response matrix; and obtain the trained multi-layer feedforward neural network surrogate model. The module for determining the optimal excitation frequency and the optimal number of PCB coil layers is configured to use the predicted value of the multilayer feedforward neural network surrogate model as the fitness function and the particle swarm optimization algorithm to search for the optimal excitation frequency and the optimal number of PCB coil layers that can produce the maximum defect response amplitude.

[0095] Example 5 The purpose of this embodiment is to provide a computer program product containing instructions that, when run on a computer, causes the computer to perform the methods and functions involved in any of the embodiments described above.

[0096] The steps and methods involved in the apparatus of the above embodiments correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0097] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0098] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for optimizing PCB eddy current detection coils based on maximizing defect response amplitude, characterized by: include: A three-dimensional electromagnetic finite element parametric simulation model was established, with the excitation frequency and the number of PCB coil layers as variables to be optimized. Based on a three-dimensional electromagnetic finite element parametric simulation model, by comparing the received signals under healthy and defective states, a defect response amplitude evaluation index is defined to characterize the probe's sensitivity to defects. The excitation frequency and the number of PCB coil layers are discretized to construct a two-dimensional parameter scanning space of frequency-layer number and perform automatic batch simulation. During the automated batch simulation process, the defect response characteristics under various parameter combinations in the frequency-layer two-dimensional parameter scanning space are calculated and a two-dimensional defect response matrix is ​​formed. The process of forming a two-dimensional defect response matrix is ​​as follows: for each set of parameters, apply the corresponding frequency excitation signal and calculate the received response under healthy conditions; Recalculate the received response under defective conditions; Calculate the defect disturbance signal; Extract the peak value of the defect response from the defect disturbance signal; After calculating all parameter combinations, a two-dimensional defect response matrix is ​​formed, where the rows correspond to different excitation frequencies and the columns correspond to different PCB layer numbers; the matrix elements represent the defect response amplitude under the corresponding parameter combination. A multi-layer feedforward neural network surrogate model is constructed with frequency and layer number as inputs and defect response amplitude as predicted output. A nonlinear mapping relationship is established, and the model is trained based on the two-dimensional defect response matrix to obtain the trained multi-layer feedforward neural network surrogate model. Using the prediction values ​​of the multilayer feedforward neural network surrogate model as the fitness function, the particle swarm optimization algorithm is used to search for the optimal excitation frequency and the optimal number of PCB coil layers that can produce the maximum defect response amplitude.

2. The PCB eddy current detection coil optimization method based on maximizing defect response amplitude as described in claim 1, characterized in that, When establishing a three-dimensional electromagnetic finite element parametric simulation model, the PCB coil is parametrically modeled, and simulation models under different frequencies and different number of layers are automatically generated. During the simulation process, an adaptive mesh generation method is used to locally refine the coil area, defect area, and near-field electromagnetic coupling area.

3. The PCB eddy current detection coil optimization method based on maximizing defect response amplitude as described in claim 1, characterized in that, When defining the defect response amplitude evaluation index, the defect disturbance signal is defined as the difference between the received signal in the defect state and the received signal in the healthy state of the three-dimensional electromagnetic finite element parameterized simulation model. The maximum value of the defect disturbance signal amplitude is then defined as the defect response amplitude evaluation index.

4. The PCB eddy current detection coil optimization method based on maximizing defect response amplitude as described in claim 1, characterized in that, it further... include: The multilayer feedforward neural network surrogate model is trained and validated to accurately predict defect response values ​​in a continuous space. Specifically, during training: The two-dimensional defect response matrix data is divided into training set, validation set and test set, and mean squared error is used as the loss function. The Adam optimization algorithm is used to update network parameters during training. Once the validation set error meets the convergence condition, the neural network training is complete.

5. The PCB eddy current detection coil optimization method based on maximizing defect response amplitude as described in claim 1, characterized in that, When using the particle swarm optimization algorithm to search for the optimal excitation frequency and the optimal number of PCB coil layers that produce the maximum defect response amplitude, the following steps are included: Using a trained neural network model, the magnitude of the defect response is predicted in a continuous parameter space; An optimization objective is established based on the optimal excitation frequency and the optimal number of PCB coil layers; The particle swarm optimization algorithm is used to perform a global optimization search on the frequency-layer parameter space: First, the position of each particle will be defined; Further define the particle's velocity vector; Subsequently, the defect response amplitude predicted by the neural network was used as the particle fitness function; After initializing the particle swarm, multiple particle positions are randomly generated within a preset frequency range and layer number range, and the following are recorded: the historical best position and the global best position for each particle. In each iteration, the particle velocity is updated based on the individual particle's optimal position and the global optimal position; Then update the particle position; The updated particle positions are then input into the neural network model and the particle fitness is recalculated. When a particle's current fitness is better than its historical best fitness, update the individual's best position; when a particle's current fitness is better than the global best fitness, update the global best position. Ultimately, the globally optimal particle position is taken as the optimization result: that is, the excitation frequency that can produce the maximum defect response amplitude is selected in combination with the PCB coil layer array as the final optimization parameters.

6. A PCB eddy current detection coil optimization system based on maximizing defect response amplitude, characterized in that, include: The simulation model building module is configured to: establish a three-dimensional electromagnetic finite element parametric simulation model, with the excitation frequency and the number of PCB coil layers as variables to be optimized; The defect response amplitude evaluation index definition module is configured to: define a defect response amplitude evaluation index to characterize the probe’s sensitivity to defects by comparing the received signals in healthy and defective states, based on a three-dimensional electromagnetic finite element parametric simulation model. The batch simulation module is configured to: discretize the excitation frequency and the number of PCB coil layers, construct a two-dimensional parameter scanning space of frequency-layer number, and automatically perform batch simulation; The two-dimensional defect response matrix forming module is configured to: calculate the defect response characteristics under each parameter combination in the frequency-layer two-dimensional parameter scan space and form a two-dimensional defect response matrix during the automatic batch simulation process; the process of forming the two-dimensional defect response matrix is: for each parameter combination, apply the corresponding frequency excitation signal and calculate the received response under the healthy state; Recalculate the received response under defective conditions; Calculate the defect disturbance signal; Extract the peak value of the defect response from the defect disturbance signal; After calculating all parameter combinations, a two-dimensional defect response matrix is ​​formed, where the rows correspond to different excitation frequencies and the columns correspond to different PCB layer numbers; the matrix elements represent the defect response amplitude under the corresponding parameter combination. The surrogate model construction module is configured to: construct a multi-layer feedforward neural network surrogate model with frequency and layer number as inputs and defect response amplitude as predicted output; establish a nonlinear mapping relationship; train the model based on the two-dimensional defect response matrix; and obtain the trained multi-layer feedforward neural network surrogate model. The module for determining the optimal excitation frequency and the optimal number of PCB coil layers is configured to use the predicted value of the multilayer feedforward neural network surrogate model as the fitness function and the particle swarm optimization algorithm to search for the optimal excitation frequency and the optimal number of PCB coil layers that can produce the maximum defect response amplitude.

7. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 5.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method described in any one of claims 1-5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it performs the steps of the method described in any one of claims 1-5.

Images