CN117669359A - Ultrasonic welding process parameter recommendation method and system for connecting sheet - Google Patents
Ultrasonic welding process parameter recommendation method and system for connecting sheet Download PDFInfo
- Publication number
- CN117669359A CN117669359A CN202311362424.8A CN202311362424A CN117669359A CN 117669359 A CN117669359 A CN 117669359A CN 202311362424 A CN202311362424 A CN 202311362424A CN 117669359 A CN117669359 A CN 117669359A
- Authority
- CN
- China
- Prior art keywords
- weight
- output
- welding
- ultrasonic welding
- connecting sheet
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003466 welding Methods 0.000 title claims abstract description 97
- 238000000034 method Methods 0.000 title claims abstract description 89
- 230000008569 process Effects 0.000 title claims abstract description 64
- 239000002245 particle Substances 0.000 claims abstract description 57
- 238000005457 optimization Methods 0.000 claims abstract description 56
- 238000012549 training Methods 0.000 claims abstract description 28
- 230000006870 function Effects 0.000 claims description 55
- 230000004913 activation Effects 0.000 claims description 32
- 230000000739 chaotic effect Effects 0.000 claims description 16
- 238000003062 neural network model Methods 0.000 claims description 15
- 238000011478 gradient descent method Methods 0.000 claims description 10
- 238000010276 construction Methods 0.000 claims description 8
- 238000013507 mapping Methods 0.000 claims description 7
- 230000003213 activating effect Effects 0.000 claims description 6
- 230000003247 decreasing effect Effects 0.000 claims description 6
- 238000013459 approach Methods 0.000 claims description 4
- 230000002708 enhancing effect Effects 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 abstract description 9
- 238000012360 testing method Methods 0.000 abstract description 2
- 210000002569 neuron Anatomy 0.000 description 6
- 230000003044 adaptive effect Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000003542 behavioural effect Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 244000144992 flock Species 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000004745 nonwoven fabric Substances 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Feedback Control In General (AREA)
Abstract
The invention relates to a method and a system for recommending technological parameters of ultrasonic welding of a connecting sheet, which belong to the technical field of ultrasonic welding of the connecting sheet and solve the problem of how to effectively optimize the technological parameters of ultrasonic welding of the connecting sheet so as to improve welding quality.A BP neural network training model is used, then an improved particle swarm optimization algorithm is used for finding out the input which enables the output of the model to be close to the target output, a chaos sequence and a self-adaptive weight strategy are introduced into the improved particle swarm optimization algorithm, global searching capacity and convergence speed of the algorithm are improved, and finally a technological parameter recommending system of ultrasonic welding of the connecting sheet is established, so that automatic optimization and recommendation of technological parameters are realized; the method can effectively optimize the process parameters, improve the welding quality, and reduce the time and cost of manual test; the invention has wide application prospect and can be applied to various occasions needing to optimize the process parameters.
Description
Technical Field
The invention belongs to the technical field of ultrasonic welding of connecting sheets, and relates to a method and a system for recommending ultrasonic welding technological parameters of the connecting sheets.
Background
Ultrasonic welding of connecting sheets is a common welding method, and technological parameters of the ultrasonic welding have important influence on welding quality. However, how to determine optimal process parameters has been a challenge.
In the existing ultrasonic welding process parameter optimization technology of the connecting sheet, a common method is to determine the optimal process parameters through a large number of experiments. This method is not only time-consuming, but also requires a lot of materials and human resources to be consumed, and is inefficient. Another common approach is to use conventional optimization algorithms, such as particle swarm optimization algorithms, to find optimal process parameters.
In the prior art, the literature (research on technological parameters and structural optimization design of ultrasonic welding systems of sanitary products production lines) (Weng Minghao) adopts an improved particle swarm algorithm (MGL-PSO) based on common leading of a plurality of global guide particles to study the welding of pp non-woven fabrics of the sanitary products production lines.
However, conventional particle swarm optimization algorithms have some problems. First, it has a slow convergence rate and requires a large number of iterations to find the optimal solution. Secondly, it is prone to be trapped in local optima, and cannot find globally optimal solutions. These problems limit the application of conventional particle swarm optimization algorithms in process parameter optimization. In addition, there is a lack of a method for predicting weld quality by a machine learning method that can effectively use historical data. This makes it impossible to fully utilize existing data in optimizing process parameters, resulting in poor optimization.
Therefore, how to improve the particle swarm optimization algorithm, improve the global searching capability and the convergence rate of the particle swarm optimization algorithm, and simultaneously predict the welding quality by using a machine learning method, so that the process parameters are effectively optimized, and the problem to be solved is currently urgent.
Disclosure of Invention
The invention aims to solve the technical problem of how to effectively optimize ultrasonic welding process parameters of a connecting sheet so as to improve welding quality.
The invention solves the technical problems through the following technical scheme:
a method for recommending ultrasonic welding technological parameters of a connecting sheet comprises the following steps:
step 1, acquiring technological parameter data and corresponding welding quality data of ultrasonic welding of a connecting sheet, and training a BP neural network model by taking the technological parameter data as an input variable and the corresponding welding quality data as a target variable so as to obtain a model for predicting welding quality;
step 2, setting a target welding quality, using a chaos sequence to enhance the global searching capability of a particle swarm optimization algorithm, using a self-adaptive weight strategy to improve the convergence rate of the particle swarm optimization algorithm, and then using an improved particle swarm optimization algorithm to find a set of recommended process parameters, wherein the set of process parameters enable the output of a model to approach the target welding quality, so that the optimization of the process parameters is realized.
Further, the method for training the BP neural network model comprises the following steps:
1) Model construction: determining the node number, initialization weight and bias of an input layer, a hidden layer and an output layer, and selecting an activation function;
2) Forward propagation: in the forward propagation stage, input data starts from an input layer, passes through nodes of each layer, and finally obtains output of an output layer, wherein the output of each layer is obtained by performing dot product operation on the input data and weight, then adding bias, and finally activating a function, and the formula is as follows:
H＝activation(X*w+b)
wherein X is input data, w is weight, b is bias, activation is activation function, and H is output;
3) Back propagation: in the back propagation stage, calculating a loss function according to the output of the output layer and the real label, and then updating the weight and the bias by a gradient descent method;
4) And (5) weight updating: updating the weight and the bias by a gradient descent method;
5) Iterative training: the process of forward propagation, backward propagation and weight updating is repeated until a preset number of iterations is reached or the value of the loss function is less than a preset threshold.
Further, the loss function adopts a mean square error, and the formula is as follows:
wherein y is pred Is a predictive tag, y true Is a true label, L is a loss function, w and b are weights and offsets,andis the derivative of the activation function.
Further, the formula for updating the weights and the biases by the gradient descent method is as follows:
where lr is the learning rate.
Further, the process parameter data includes: welding time, welding power, pre-welding height, post-welding height, pre-and post-welding height difference, welding energy, welding amplitude, and welding pressure.
Further, the chaotic sequence is generated by using Logistic mapping, and the specific formula is as follows:
x id (t+1)＝μ*x id (t)*(1-x id (t))
wherein μ is a control parameter, x id (t+1) and x id (t) is the position of the particle i at the t+1 generation and the t generation, respectively.
Further, the adaptive weight strategy adopts a linear decreasing weight strategy, and the specific formula is as follows:
w(t)＝w max -(w max -w min )*t/t max
where t is the current iteration number, w (t) is the weight of the current iteration number, w max And w min Respectively the maximum value and the minimum value of the weight, t max Is the maximum number of iterations.
A technological parameter recommendation system for ultrasonic welding of connecting sheets comprises: the system comprises a data acquisition and model training module and a process parameter optimization module;
the data acquisition and model training module is used for acquiring technological parameter data and corresponding welding quality data of ultrasonic welding of the connecting sheet, and training a BP neural network model by taking the technological parameter data as an input variable and the corresponding welding quality data as a target variable so as to obtain a model for predicting welding quality;
the process parameter optimization module is used for setting a target welding quality, enhancing the global searching capability of the particle swarm optimization algorithm by using a chaotic sequence, improving the convergence rate of the particle swarm optimization algorithm by using a self-adaptive weight strategy, and searching a set of recommended process parameters by using the improved particle swarm optimization algorithm, wherein the set of process parameters enable the output of the model to be close to the target welding quality, so that the optimization of the process parameters is realized.
Further, the method for training the BP neural network model comprises the following steps:
1) Model construction: determining the node number, initialization weight and bias of an input layer, a hidden layer and an output layer, and selecting an activation function;
2) Forward propagation: in the forward propagation stage, input data starts from an input layer, passes through nodes of each layer, and finally obtains output of an output layer, wherein the output of each layer is obtained by performing dot product operation on the input data and weight, then adding bias, and finally activating a function, and the formula is as follows:
H＝activation(X*w+b)
wherein X is input data, w is weight, b is bias, activation is activation function, and H is output;
3) Back propagation: in the back propagation stage, calculating a loss function according to the output of the output layer and the real label, and then updating the weight and the bias by a gradient descent method;
the loss function adopts a mean square error, and the formula is as follows:
wherein y is pred Is a predictive tag, y true Is a true label, L is a loss function, w and b are weights and offsets,andis the derivative of the activation function;
4) And (5) weight updating: the weights and offsets are updated by gradient descent, the formula is as follows:
where lr is the learning rate;
5) Iterative training: the process of forward propagation, backward propagation and weight updating is repeated until a preset number of iterations is reached or the value of the loss function is less than a preset threshold.
Further, the chaotic sequence is generated by using Logistic mapping, and the specific formula is as follows:
x id (t+1)＝μ*x id (t)*(1-x id (t))
wherein μ is a control parameter, x id (t+1) and x id (t) the positions of the particles i at the t+1 generation and the t generation, respectively;
the self-adaptive weight strategy adopts a linear decreasing weight strategy, and the specific formula is as follows:
w(t)＝w max -(w max -w min )*t/t max
where t is the current iteration number, w (t) is the weight of the current iteration number, w max And w min Respectively the maximum value and the minimum value of the weight, t max Is the maximum number of iterations.
The invention has the advantages that:
according to the invention, a BP neural network is used for training a model, then an improved particle swarm optimization algorithm is used for finding an input enabling the output of the model to approach to the target output, a chaotic sequence and a self-adaptive weight strategy are introduced into the improved particle swarm optimization algorithm, global searching capacity and convergence speed of the algorithm are improved, and finally a connecting piece ultrasonic welding process parameter recommendation system is established, so that automatic optimization and recommendation of process parameters are realized; the method can effectively optimize the process parameters, improve the welding quality, and reduce the time and cost of manual test; the invention has wide application prospect and can be applied to various occasions needing to optimize the process parameters.
Drawings
FIG. 1 is a flowchart of a method for recommending ultrasonic welding process parameters for a connecting sheet according to an embodiment of the invention;
FIG. 2 is an intelligent recommended input variable for setting the pull force magnitude to 250N for a polar cathode;
fig. 3 is an intelligent recommended input variable for setting the pull force to 300N for the positive polarity.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions in the embodiments of the present invention will be clearly and completely described in the following in conjunction with the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The technical scheme of the invention is further described below with reference to the attached drawings and specific embodiments:
example 1
As shown in fig. 1, the method for recommending ultrasonic welding process parameters of the connecting sheet according to the embodiment of the invention comprises the following contents:
1. BP neural network training model
Firstly, collecting technological parameter data of ultrasonic welding of a batch of connecting sheets, such as welding time, welding power, pre-welding height, post-welding height, pre-welding and post-welding height difference, welding energy, welding amplitude, welding pressure and corresponding welding quality data, taking the technological parameter data as an input variable, taking the corresponding welding quality data as a target variable, and training a BP neural network model by using the data to train a model capable of predicting welding quality.
The BP neural network, i.e. the counter-propagating neural network (Backpropagation Neural Network), the basic structure comprises three layers: input layer, hidden layer and output layer. The input layer accepts input data, and the hidden layer and the output layer are composed of a plurality of neurons. Each neuron is connected to each neuron of the previous layer and has a weight. As the input data passes through the neural network, each neuron sums the weighted inputs and performs a nonlinear transformation through an activation function, and then passes the output to the next layer of neurons. Thus, after layer-by-layer delivery, the final output layer will produce an output result.
The training process of the BP neural network is based on an error back propagation algorithm (Backpropagation), i.e. the weight value of each neuron is updated by back propagation of the error to reduce the error between the network output and the real output. The objective function of the error calculation is typically a square error or a cross entropy error.
The following is the specific implementation principle and mathematical logic of training data and model construction of the BP neural network:
(1) Model construction: the construction of the BP neural network model mainly comprises the steps of determining a network structure (the node numbers of an input layer, a hidden layer and an output layer), initializing weights, biasing, selecting an activation function and the like. The hidden layer of the BP neural network model used in the invention is 11 layers, and the used activation function is a ReLU function and the like.
(2) Forward propagation: in the forward propagation phase, input data starts from the input layer and passes through the nodes of each layer, and finally output of the output layer is obtained. The output of each layer is obtained by performing dot product operation on the input data and the weights, then adding bias, and finally activating the function. The formula is as follows:
H＝activation(X*w+b)
where X is the input data, w is the weight, b is the bias, activation is the activation function, and H is the output.
(3) Back propagation: in the back propagation phase, a loss function is calculated from the output of the output layer and the real labels, and then the weights and biases are updated by a gradient descent method. The loss function used in the present invention is the Mean Square Error (MSE). The formula is as follows:
wherein y is pred Is a predictive tag, y true Is a true label, L is a loss function, w and b are weights and offsets,andis the derivative of the activation function.
(4) And (5) weight updating: the weights and biases are updated by gradient descent. The formula is as follows:
where lr is the learning rate, lr used in the present invention is 0.001.
(5) Iterative training: the process of forward propagation, backward propagation and weight updating is repeated until a preset number of iterations is reached or the value of the loss function is less than a preset threshold.
2. Improved particle swarm optimization algorithm
And introducing a chaotic sequence and an adaptive weight strategy into the particle swarm optimization algorithm. The chaotic sequence has good randomness and ergodic property, and can enhance the global searching capability of the algorithm; the self-adaptive weight strategy can dynamically adjust the weight according to the change of the searching process, so that the convergence rate of the algorithm is improved.
The specific mathematical logic and formula is as follows:
(1) Particle swarm optimization algorithm
The particle swarm optimization algorithm (PSO: particle swarm optimization) is an evolutionary computing technique (evolutionary computation). From behavioral studies on flock predation. Basic idea of particle swarm optimization algorithm: the optimal solution is found through collaboration and information sharing among individuals in the population.
Particle swarm algorithm simulates birds in a bird swarm by designing a mass-free particle that has only two properties: speed, which represents the speed of movement, and position, which represents the direction of movement. Each particle independently searches for an optimal solution in a search space, marks the optimal solution as a current individual extremum, shares the individual extremum with other particles in the whole particle swarm, finds the optimal individual extremum as a current global optimal solution of the whole particle swarm, and adjusts the speed and the position of each particle in the particle swarm according to the current individual extremum found by each particle and the current global optimal solution shared by the whole particle swarm.
In the standard PSO algorithm, the particles find the optimal particles by learning their own historical experience (pbest) and population experience (gbest). For the solution variable, x= { X 1 ,x 2 ,…,x D The optimization problem of min { f (x) } as an objective function, and the standard PSO algorithm particle update formula is as follows:
v id (t+1)＝wv id (t)+c 1 r 1 (pbest id -x id (t))+c 2 r 2 (gbest d -x id (t))x id (t+1)＝x id (t)+v id (t+1)
in the formula, v id (t+1) and x id (t+1) is the velocity and position of the particle i at the t+1 generation, respectively; w is an inertial weight, and w linearly decreases with iteration times in a standard PSO algorithm; c 1 And c 2 Is a learning factor; r is (r) 1 And r 2 Is [0,1 ]]Random numbers distributed uniformly.
(2) Introduction of chaotic sequence for improved algorithm
In the particle swarm optimization algorithm, the position of each particle represents one possible solution, and the velocity represents the direction and magnitude of the change in solution. Chaotic sequences are introduced in the present invention to enhance the global search capability of the algorithm. The chaotic sequence is a non-periodic, deterministic random sequence generated by chaotic mapping. In the invention, a Logistic mapping is used for generating a chaotic sequence, and the specific formula is as follows:
x id (t+1)＝μ*x id (t)*(1-x id (t))
wherein μ is a control parameterThe value range is [0,4 ]]。x id (t+1) and x id (t) is the position of the particle i at the t+1 generation and the t generation, respectively.
(3) Improved algorithm introduction adaptive weight strategy
The self-adaptive weight strategy is to dynamically adjust the weight according to the change of the searching process, so that the convergence rate of the algorithm is improved. In the invention, a linear decreasing weight strategy is used, and the specific formula is as follows:
w(t)＝w max -(w max -w min )*t/t maax
where t is the current iteration number, w (t) is the weight of the current iteration number, w max And w min Respectively the maximum value and the minimum value of the weight, t max Is the maximum number of iterations.
3. Finding an input that approximates the model output to the target output using an improved particle swarm optimization algorithm
And (3) by improving a particle swarm optimization algorithm, the process parameter which enables the BP neural network model output to be close to the target output is found, so that the optimization of the process parameter is realized.
The specific mathematical logic and formula is as follows:
minf(x)＝||y-y′||
wherein x is a process parameter, y is the output of the BP neural network model, y' is the actual target output, i·i is the Euclidean distance, f (x) is an optimization objective function, and the optimization objective function in the invention is the improved particle swarm optimization algorithm.
Example two
A technological parameter recommendation system for ultrasonic welding of connecting sheets comprises: the system comprises a data acquisition and model training module and a process parameter optimization module;
the data acquisition and model training module is used for acquiring technological parameter data and corresponding welding quality data of ultrasonic welding of the connecting sheet, and training a BP neural network model by taking the technological parameter data as an input variable and the corresponding welding quality data as a target variable so as to obtain a model for predicting welding quality;
the process parameter optimization module is used for setting a target welding quality, enhancing the global searching capability of the particle swarm optimization algorithm by using a chaotic sequence, improving the convergence rate of the particle swarm optimization algorithm by using a self-adaptive weight strategy, and searching a set of recommended process parameters by using the improved particle swarm optimization algorithm, wherein the set of process parameters enable the output of the model to be close to the target welding quality, so that the optimization of the process parameters is realized.
The chaotic sequence is generated by using Logistic mapping, and the specific formula is as follows:
x id (t+1)＝μ*x id (t)*(1-x id (t))
wherein μ is a control parameter, x id (t+1) and x id (t) the positions of the particles i at the t+1 generation and the t generation, respectively;
the self-adaptive weight strategy adopts a linear decreasing weight strategy, and the specific formula is as follows:
w(t)＝w max -(w max -w min )*t/t max
where t is the current iteration number, w (t) is the weight of the current iteration number, w max And w min Respectively the maximum value and the minimum value of the weight, t max Is the maximum number of iterations.
The method for training the BP neural network model comprises the following steps:
1) Model construction: determining the node number, initialization weight and bias of an input layer, a hidden layer and an output layer, and selecting an activation function;
2) Forward propagation: in the forward propagation stage, input data starts from an input layer, passes through nodes of each layer, and finally obtains output of an output layer, wherein the output of each layer is obtained by performing dot product operation on the input data and weight, then adding bias, and finally activating a function, and the formula is as follows:
H＝activation(X*w+b)
wherein X is input data, w is weight, b is bias, activation is activation function, and H is output;
3) Back propagation: in the back propagation stage, calculating a loss function according to the output of the output layer and the real label, and then updating the weight and the bias by a gradient descent method;
the loss function adopts a mean square error, and the formula is as follows:
wherein y is pred Is a predictive tag, y true Is a true label, L is a loss function, w and b are weights and offsets,andis the derivative of the activation function;
4) And (5) weight updating: the weights and offsets are updated by gradient descent, the formula is as follows:
where lr is the learning rate;
5) Iterative training: the process of forward propagation, backward propagation and weight updating is repeated until a preset number of iterations is reached or the value of the loss function is less than a preset threshold.
As shown in fig. 2 and 3, the Web application of the intelligent recommended technological parameters of the ultrasonic welding process of the connecting sheet and the system of the invention allows the user to select the polarity and set the drawing force target, fig. 2 is an intelligent recommended input variable when the drawing force is set to 250N and the polarity is negative, fig. 3 is an intelligent recommended input variable when the drawing force is set to 300N and the polarity is positive, and the user intelligently recommends proper technological parameters according to the selected polarity and the target drawing force by the background prediction model.
The above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. The ultrasonic welding process parameter recommendation method for the connecting sheet is characterized by comprising the following steps of:
step 1, acquiring technological parameter data and corresponding welding quality data of ultrasonic welding of a connecting sheet, and training a BP neural network model by taking the technological parameter data as an input variable and the corresponding welding quality data as a target variable so as to obtain a model for predicting welding quality;
step 2, setting a target welding quality, using a chaos sequence to enhance the global searching capability of a particle swarm optimization algorithm, using a self-adaptive weight strategy to improve the convergence rate of the particle swarm optimization algorithm, and then using an improved particle swarm optimization algorithm to find a set of recommended process parameters, wherein the set of process parameters enable the output of a model to approach the target welding quality, so that the optimization of the process parameters is realized.
2. The ultrasonic welding process parameter recommendation method for the connecting sheet according to claim 1, wherein the method for training the BP neural network model is as follows:
1) Model construction: determining the node number, initialization weight and bias of an input layer, a hidden layer and an output layer, and selecting an activation function;
2) Forward propagation: in the forward propagation stage, input data starts from an input layer, passes through nodes of each layer, and finally obtains output of an output layer, wherein the output of each layer is obtained by performing dot product operation on the input data and weight, then adding bias, and finally activating a function, and the formula is as follows:
H＝activation(X*+)
wherein X is input data, w is weight, b is bias, activation is activation function, and H is output;
3) Back propagation: in the back propagation stage, calculating a loss function according to the output of the output layer and the real label, and then updating the weight and the bias by a gradient descent method;
4) And (5) weight updating: updating the weight and the bias by a gradient descent method;
5) Iterative training: the process of forward propagation, backward propagation and weight updating is repeated until a preset number of iterations is reached or the value of the loss function is less than a preset threshold.
3. The ultrasonic welding process parameter recommendation method for the connecting sheet according to claim 2, wherein the loss function adopts a mean square error as follows:
wherein y is pred Is a predictive tag, y true Is a true label, L is a loss function, w and b are weights and biasThe device is arranged in the way that the device is arranged,and->Is the derivative of the activation function.
4. The ultrasonic welding process parameter recommendation method for the connecting sheet according to claim 3, wherein the formula for updating the weight and the bias by the gradient descent method is as follows:
where lr is the learning rate.
5. The ultrasonic welding process parameter recommendation method for the connecting sheet according to claim 1, wherein the process parameter data comprises: welding time, welding power, pre-welding height, post-welding height, pre-and post-welding height difference, welding energy, welding amplitude, and welding pressure.
6. The ultrasonic welding process parameter recommendation method for the connecting sheet according to claim 1, wherein the chaotic sequence is generated by using Logistic mapping, and the specific formula is as follows:
x id (t+1)＝μ*x id (t)*(1-x id (t))
wherein μ is a control parameter, x id (t+1) and x id (t) is the position of the particle i at the t+1 generation and the t generation, respectively.
7. The ultrasonic welding process parameter recommendation method for the connecting sheet according to claim 1, wherein the self-adaptive weight strategy adopts a linear decreasing weight strategy, and the specific formula is as follows:
w(t)＝w max -(w max -w min )*t/t max
where t is the current iteration number, w (t) is the weight of the current iteration number, w max And w min Respectively the maximum value and the minimum value of the weight, t max Is the maximum number of iterations.
8. A connection piece ultrasonic welding process parameter recommendation system is characterized by comprising: the system comprises a data acquisition and model training module and a process parameter optimization module;
the data acquisition and model training module is used for acquiring technological parameter data and corresponding welding quality data of ultrasonic welding of the connecting sheet, and training a BP neural network model by taking the technological parameter data as an input variable and the corresponding welding quality data as a target variable so as to obtain a model for predicting welding quality;
the process parameter optimization module is used for setting a target welding quality, enhancing the global searching capability of the particle swarm optimization algorithm by using a chaotic sequence, improving the convergence rate of the particle swarm optimization algorithm by using a self-adaptive weight strategy, and searching a set of recommended process parameters by using the improved particle swarm optimization algorithm, wherein the set of process parameters enable the output of the model to be close to the target welding quality, so that the optimization of the process parameters is realized.
9. The ultrasonic welding process parameter recommendation system for connecting sheets according to claim 8, wherein the method for training the BP neural network model is as follows:
1) Model construction: determining the node number, initialization weight and bias of an input layer, a hidden layer and an output layer, and selecting an activation function;
2) Forward propagation: in the forward propagation stage, input data starts from an input layer, passes through nodes of each layer, and finally obtains output of an output layer, wherein the output of each layer is obtained by performing dot product operation on the input data and weight, then adding bias, and finally activating a function, and the formula is as follows:
H＝activation(X*+)
wherein X is input data, w is weight, b is bias, activation is activation function, and H is output;
3) Back propagation: in the back propagation stage, calculating a loss function according to the output of the output layer and the real label, and then updating the weight and the bias by a gradient descent method;
the loss function adopts a mean square error, and the formula is as follows:
wherein y is pred Is a predictive tag, y true Is a true label, L is a loss function, w and b are weights and offsets,and->Is the derivative of the activation function;
4) And (5) weight updating: the weights and offsets are updated by gradient descent, the formula is as follows:
where lr is the learning rate;
5) Iterative training: the process of forward propagation, backward propagation and weight updating is repeated until a preset number of iterations is reached or the value of the loss function is less than a preset threshold.
10. The ultrasonic welding process parameter recommendation system for the connecting sheet according to claim 8, wherein the chaotic sequence is generated by using a Logistic mapping, and the specific formula is as follows:
x id (+1)＝μ* id ()*(1-x id ())
wherein μ is a control parameter, x id (+1) and x id () The positions of the particles i at the t+1 generation and the t generation respectively;
the self-adaptive weight strategy adopts a linear decreasing weight strategy, and the specific formula is as follows:
w(t)＝w max -(w max -w min )*t/t max
where t is the current iteration number, w (t) is the weight of the current iteration number, w max And w min Respectively the maximum value and the minimum value of the weight, t max Is the maximum number of iterations.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311362424.8A CN117669359A (en) | 2023-10-19 | 2023-10-19 | Ultrasonic welding process parameter recommendation method and system for connecting sheet |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311362424.8A CN117669359A (en) | 2023-10-19 | 2023-10-19 | Ultrasonic welding process parameter recommendation method and system for connecting sheet |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117669359A true CN117669359A (en) | 2024-03-08 |
Family
ID=90074114
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311362424.8A Pending CN117669359A (en) | 2023-10-19 | 2023-10-19 | Ultrasonic welding process parameter recommendation method and system for connecting sheet |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117669359A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118132849A (en) * | 2024-03-22 | 2024-06-04 | 青岛铭品饰家装饰工程有限公司 | Steel structure welding process quality management recommendation method based on big data processing |
-
2023
- 2023-10-19 CN CN202311362424.8A patent/CN117669359A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118132849A (en) * | 2024-03-22 | 2024-06-04 | 青岛铭品饰家装饰工程有限公司 | Steel structure welding process quality management recommendation method based on big data processing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Smith et al. | Evolutionary multi-objective generation of recurrent neural network ensembles for time series prediction | |
| Ren et al. | Optimal parameters selection for BP neural network based on particle swarm optimization: A case study of wind speed forecasting | |
| CN111191934B (en) | Multi-target cloud workflow scheduling method based on reinforcement learning strategy | |
| Han et al. | Adaptive computation algorithm for RBF neural network | |
| Ge et al. | Multi-agent transfer reinforcement learning with multi-view encoder for adaptive traffic signal control | |
| Wang et al. | Fixed-time synchronization for complex-valued BAM neural networks with time-varying delays via pinning control and adaptive pinning control | |
| CN111415051A (en) | Power material demand plan prediction method based on BP neural network | |
| Barzola-Monteses et al. | Energy consumption of a building by using long short-term memory network: a forecasting study | |
| CN104050505A (en) | Multilayer-perceptron training method based on bee colony algorithm with learning factor | |
| CN117669359A (en) | Ultrasonic welding process parameter recommendation method and system for connecting sheet | |
| Wang et al. | A novel median dendritic neuron model for prediction | |
| Ji et al. | Forecasting wind speed time series via dendritic neural regression | |
| CN113141012A (en) | Power grid power flow regulation and control decision reasoning method based on deep deterministic strategy gradient network | |
| Dong et al. | Lambo: Large language model empowered edge intelligence | |
| Na et al. | A novel heuristic artificial neural network model for urban computing | |
| Lucas et al. | Methodology based on spiking neural networks for univariate time-series forecasting | |
| CN114154688A (en) | Short-term power prediction method for photovoltaic power station | |
| Khosrowshahi | Innovation in artificial neural network learning: Learn-On-Demand methodology | |
| Li | Research on bank credit risk assessment based on BP neural network | |
| Rozario et al. | In-Depth Case Study on Artificial Neural Network Weights Optimization Using Meta-Heuristic and Heuristic Algorithmic Approach | |
| Sulaiman et al. | Evolutionary programming versus artificial immune system in evolving neural network for grid-connected photovoltaic system output prediction | |
| Chen et al. | Recurrent fuzzy system design using mutation-aided elite continuous ant colony optimization | |
| Venayagamoorthy | Teaching neural networks concepts and their learning techniques | |
| Zhang et al. | Sunspot activity prediction based on adaptive hybrid algorithms | |
| Zhang et al. | Task Assignment of Multi-AUVs Based on Artificial Bee Colony Genetic Hybrid Algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |