CN114980527A - Circuit board soldering path planning method and device and electronic equipment - Google Patents
Circuit board soldering path planning method and device and electronic equipment Download PDFInfo
- Publication number
- CN114980527A CN114980527A CN202210725876.7A CN202210725876A CN114980527A CN 114980527 A CN114980527 A CN 114980527A CN 202210725876 A CN202210725876 A CN 202210725876A CN 114980527 A CN114980527 A CN 114980527A
- Authority
- CN
- China
- Prior art keywords
- circuit board
- soldering
- pointer network
- point
- points
- 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.)
- Granted
Links
- 238000005476 soldering Methods 0.000 title claims abstract description 124
- 238000000034 method Methods 0.000 title claims abstract description 49
- 238000003466 welding Methods 0.000 claims abstract description 39
- 230000008569 process Effects 0.000 claims abstract description 14
- ATJFFYVFTNAWJD-UHFFFAOYSA-N Tin Chemical compound [Sn] ATJFFYVFTNAWJD-UHFFFAOYSA-N 0.000 claims description 20
- 238000004422 calculation algorithm Methods 0.000 claims description 20
- 229910000679 solder Inorganic materials 0.000 claims description 7
- 230000002787 reinforcement Effects 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 3
- 238000004590 computer program Methods 0.000 claims description 2
- 238000005457 optimization Methods 0.000 abstract description 7
- 238000013528 artificial neural network Methods 0.000 abstract description 4
- 239000000203 mixture Substances 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 6
- 239000013598 vector Substances 0.000 description 6
- 238000013135 deep learning Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000004519 manufacturing process Methods 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
Images
Classifications
-
- H—ELECTRICITY
- H05—ELECTRIC TECHNIQUES NOT OTHERWISE PROVIDED FOR
- H05K—PRINTED CIRCUITS; CASINGS OR CONSTRUCTIONAL DETAILS OF ELECTRIC APPARATUS; MANUFACTURE OF ASSEMBLAGES OF ELECTRICAL COMPONENTS
- H05K3/00—Apparatus or processes for manufacturing printed circuits
- H05K3/0005—Apparatus or processes for manufacturing printed circuits for designing circuits by computer
-
- H—ELECTRICITY
- H05—ELECTRIC TECHNIQUES NOT OTHERWISE PROVIDED FOR
- H05K—PRINTED CIRCUITS; CASINGS OR CONSTRUCTIONAL DETAILS OF ELECTRIC APPARATUS; MANUFACTURE OF ASSEMBLAGES OF ELECTRICAL COMPONENTS
- H05K3/00—Apparatus or processes for manufacturing printed circuits
- H05K3/0002—Apparatus or processes for manufacturing printed circuits for manufacturing artworks for printed circuits
-
- H—ELECTRICITY
- H05—ELECTRIC TECHNIQUES NOT OTHERWISE PROVIDED FOR
- H05K—PRINTED CIRCUITS; CASINGS OR CONSTRUCTIONAL DETAILS OF ELECTRIC APPARATUS; MANUFACTURE OF ASSEMBLAGES OF ELECTRICAL COMPONENTS
- H05K3/00—Apparatus or processes for manufacturing printed circuits
- H05K3/30—Assembling printed circuits with electric components, e.g. with resistor
- H05K3/32—Assembling printed circuits with electric components, e.g. with resistor electrically connecting electric components or wires to printed circuits
- H05K3/328—Assembling printed circuits with electric components, e.g. with resistor electrically connecting electric components or wires to printed circuits by welding
-
- 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)
- Manufacturing & Machinery (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Computer Hardware Design (AREA)
- Electric Connection Of Electric Components To Printed Circuits (AREA)
Abstract
The application is suitable for the technical field of circuit board soldering, and provides a path planning method and device for circuit board soldering and electronic equipment, wherein the method comprises the following steps: and extracting the soldering point coordinates by using pixel differences in the same circuit board image to obtain a soldering point set, dividing confidence intervals in the soldering point set according to the real circuit board label, carrying out forward propagation by using a pointer network, carrying out end-to-end processing on an input sample sequence, and generating a circuit board soldering path. In the training process of the pointer network, a strategy of continuous welding of pins in the same confidence space is added, and classified welding of elements on the same circuit board can be realized. Because the pointer network is a neural network for solving the regression task essentially, and the optimization methods of circuit boards with various compositions are learned from a large number of samples during training, compared with the traditional optimization method, the generalization is better, the local optimal solution is not easy to fall into, and the planning part only needs to realize forward propagation once, so that the speed is higher.
Description
Technical Field
The application belongs to the technical field of circuit board soldering, and particularly relates to a path planning method and device for circuit board soldering and electronic equipment.
Background
With the surface area of the components on the PCB being smaller and smaller, the pins of the components are denser and denser, and the number of welding points is larger and larger, so that the detection of the soldering points becomes more and more difficult. At present, the positioning of the position of a soldering point by using an image processing technology is a commonly used technical means, and how to sequence the soldering sequence of the soldering point in a centralized coordinate point of the central position of the soldering point on a circuit board and realize the optimal welding strategy is a problem worthy of discussion.
The existing welding sequence planning algorithm for the PCB soldering points usually uses heuristic algorithms, and the planning algorithms extract the coordinates of the PCB soldering points and process the extracted coordinates as a traveling salesman model (TSP), namely, the problem that a complete graph of n points is given, each side has a length, and a closed loop which has the shortest total length and passes through each vertex exactly once is solved. That is, these algorithms do not consider that there are many components on a single circuit board, these components often correspond to different soldering processes during the assembly process, and in the real production environment, in various complicated and varied circuit board models, only rely on the manually-made heuristic guidance search process to effectively search for the optimal solution, which may be accompanied by the variation of the circuit board and easily fall into the local optimal solution, and at this time, the algorithm needs to be adjusted, and the generalization is poor.
Disclosure of Invention
The application aims to provide a path planning method and device for circuit board soldering and electronic equipment, and aims to solve the technical problems that the planning of the circuit board soldering path in the prior art does not have the capability of identifying confidence spaces corresponding to different elements and can be in a local optimal solution.
In a first aspect, the present application provides a method for planning a path for soldering a circuit board, the method including:
training a pointer network model by using a pointer network as a path planning backbone network;
and planning paths of the welding point sets on the soldered circuit board by adopting the trained pointer network model to generate a soldered path of the circuit board.
The technical scheme adopted by the embodiment of the application further comprises the following steps: when the pointer network is used as a path planning main network for training a pointer network model, a reinforcement learning algorithm Actor-criticic is used for carrying out gradient descent on the pointer network of the main network, and in the descending process, the following weighted distance formula is adopted to modify the loss function of the criticic network:
the technical scheme adopted by the embodiment of the application further comprises the following steps: when the pointer network is used as a path planning backbone network for training a pointer network model, the method further comprises the following steps:
acquiring sample images of more than ten thousand levels;
adding a component pin point set into the sample image;
dividing confidence intervals in the sample image according to the component pin point set;
and calculating the distance between the point sets by adopting the weighted distance summation formula in the confidence space.
The technical scheme adopted by the embodiment of the application further comprises the following steps: when a pointer network model is adopted to plan the path of a welding spot set on a soldering circuit board,
extracting a welding spot coordinate point set on the soldering circuit board;
and taking the welding point coordinate point set as input, and carrying out soldering path planning by adopting the network parameter forward propagation in the pointer network model after training.
The technical scheme adopted by the embodiment of the application further comprises the following steps: the step of extracting the welding spot coordinate point set of the soldering circuit board comprises the following steps:
extracting the numerical value of a pixel point in the circuit board image;
extracting a central coordinate of which the pixel point value is approximate to the pixel value of a real welding point from the circuit board image;
and determining the center coordinates as corresponding welding point coordinates.
The technical scheme adopted by the embodiment of the application further comprises the following steps: in generating the training samples:
generating part of simulated soldering points according to the layout of pins of different elements;
randomly generating partial soldering points;
and generating a training sample according to the soldering points.
The technical scheme adopted by the embodiment of the application further comprises the following steps: the step of adopting a pre-trained pointer network model to forward propagate the soldering point sets for dividing confidence intervals to generate a circuit board soldering path comprises the following steps:
the technical scheme adopted by the embodiment of the application further comprises the following steps: the step of extracting the soldering points from the circuit board image to obtain the soldering point coordinate point set according to the numerical difference value of each pixel point in the circuit board image comprises the following steps:
graying the circuit board image to obtain a circuit board grayscale image;
and extracting the coordinates of the center position of the soldering point from the gray image of the circuit board according to the gray value of each pixel point in the gray image of the circuit board to obtain a soldering point coordinate point set.
The technical scheme adopted by the embodiment of the application further comprises the following steps: the step of adopting a pre-trained pointer network model to forward propagate the soldering point sets for dividing confidence intervals to generate a circuit board soldering path comprises the following steps:
aiming at different tin soldering points, a confidence space is divided for the belonging element of the welding points on the circuit board to the welding point set;
and continuously welding the welding points under the same identified confidence space on the circuit board of the pointer network model trained in advance.
The technical scheme adopted by the embodiment of the application further comprises the following steps:
generating a point set training set by using a program;
and training the training set by adopting a deep reinforcement learning algorithm to generate a pointer network model.
In a second aspect, the present application further provides a path planning apparatus for circuit board soldering, including:
the tin soldering point extraction module is used for extracting tin soldering points from the circuit board image according to the numerical values of all pixel points in the circuit board image to obtain a tin soldering point set;
the confidence interval division module is used for dividing confidence intervals in the tin soldering point set according to the attribution elements of the tin soldering points in the tin soldering point set;
and the soldering path generation module is used for adopting a pre-trained pointer network model to forward propagate the soldering point sets for dividing the confidence intervals to generate the circuit board soldering path.
In a third aspect, the present application further provides an electronic device, including:
a processor; and
a memory communicatively coupled to the processor; wherein the content of the first and second substances,
the memory stores readable instructions which, when executed by the processor, implement the method of the first aspect.
In a fourth aspect, the present application provides a computer readable storage medium having stored thereon a computer program which, when executed, implements the method of the first aspect.
Compared with the prior art, the embodiment of the application has the advantages that: in the method, the device and the electronic equipment for circuit board soldering path planning, after soldering coordinate points are extracted from a circuit board image according to values of all pixel points in the circuit board image to obtain point sets, confidence intervals can be automatically divided in the soldering point sets according to the tin soldering point sets belonging elements, and then the tin soldering point sets dividing the confidence intervals are subjected to forward propagation by adopting a pre-trained pointer network model to generate a circuit board soldering path. The method classifies soldering points according to confidence intervals of different elements in a circuit board image, so that pin soldering points of the same element are planned preferentially, continuity of a soldering sequence of pins of the same element is fully considered, a pointer network model algorithm is essentially a neural network for solving a regression task, a large number of circuit board training samples are learned during network training, and after the network training is finished, forward propagation is needed only once without an optimization process like a traditional optimization algorithm, so that the generalization is better, and the method can be suitable for different circuit boards.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:
fig. 1 is a flowchart illustrating an implementation of a path planning method for circuit board soldering according to an embodiment.
Fig. 2 is a schematic diagram illustrating a circuit board image according to an embodiment.
FIG. 3 is a diagram illustrating a point set training process, according to an embodiment.
Fig. 4 is a block diagram of a path planning apparatus for circuit board soldering according to the second embodiment.
Detailed Description
In order to make the technical solutions better understood by those skilled in the art, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only partial embodiments of the present application, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
It should be noted that the terms "first," "second," and the like in the description and claims of this application and in the accompanying drawings are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the application described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
Specific implementations of the present application are described in detail below with reference to specific embodiments:
the first embodiment is as follows:
fig. 1 is a flowchart illustrating an implementation of a path planning method for circuit board soldering according to an embodiment. The path planning method for circuit board soldering shown in the first embodiment is applicable to electronic equipment, and a processor is arranged in the electronic equipment to realize path planning for circuit board soldering, so that the circuit board soldering path has continuity and generalization for the same component.
And S100, training a pointer network model by using the pointer network as a path planning backbone network.
And S200, performing path planning on a welding spot set on the soldered circuit board by adopting the trained pointer network model to generate a soldered path of the circuit board.
The training pointer network can identify different elements in the circuit board and divide the confidence space for the pins of the elements, so that the pins of the elements in different confidence spaces can be classified and welded, and the subsequent process optimization is served.
Specifically, the circuit board soldering path can be generated in the following manner.
And step S110, extracting the position coordinates of the soldering points from the circuit board image of the electronic equipment according to the numerical values of all the pixel points in the circuit board image to obtain a soldering point set.
And step S120, according to the belonging elements of the soldering points in the soldering point set of the electronic equipment, planning a part of pointer networks to divide confidence intervals in the soldering point set of the electronic equipment.
And S130, using a pointer network model of the planning part to forward propagate the soldering point sets of the electronic equipment and the confidence intervals, and generating a circuit board soldering path.
The solder point set is a set including information on coordinates of solder point positions.
It should be noted that, when the tin soldering point is extracted according to each pixel point in the circuit board image, the tin soldering point can be extracted according to the color value of the pixel point, and can also be extracted according to other types of values of the pixel point.
For example, the coordinates of the center of the pixel value region can be easily and directly obtained by using the opencv library according to the value difference of the pixel.
Common PCB circuit boards are mostly green oil boards and black oil boards, and soldering points generally have larger color difference with the circuit boards.
As shown in fig. 2, the color of the solder point is clearly different from the colors of other positions in the circuit board image. Therefore, the soldering points can be extracted from the circuit board image according to the color of each pixel point in the circuit board image. Specifically, when the soldering points are extracted from the electronic device circuit board image according to the color of each pixel point in the circuit board image, the circuit board image can be grayed to obtain a circuit board gray image, and then the soldering points are extracted from the electronic device circuit board gray image according to the gray values of each pixel point in the electronic device circuit board gray image to obtain a soldering point set.
For example, firstly, a set of program for extracting coordinates of a welding point of a circuit board is designed, and a circuit board image is firstly used as the input of the program and is converted into a gray image:
Grey＝0.299*R+0.587*G+0.114*B
gray is a gray value, R is a red value (255, 0, 0) in an RGB image, G is a green value (0, 255, 0) in an image, and B is a blue value (0, 0, 255) in an RGB image.
And traversing in the circuit board image, finding out pixel points with the gray values approximate to the gray values of the welding points, and taking the center coordinates of the pixel points as the positions of the soldering points of the PCB.
A large number of components such as single-chip microcomputers, chips and the like exist on the circuit board, pins of the components are distributed densely and symmetrically, and the confidence interval is the interval of soldering points of the pins.
Therefore, confidence intervals are divided in the circuit board image through the pointer network, all soldering points can be classified, pin soldering points of the same component are planned in priority, and continuity of a pin soldering sequence of the same component is fully considered.
Specifically, during pointer network training, a portion of the training set is added to a set of solder points similar to the component pins, and the distances in this set of points are weighted distances in order for the pointer network to learn to identify the confidence space.
After the confidence intervals are divided, forward propagation operation is carried out on the soldering point set by adopting a pre-trained pointer network model, and a circuit board soldering path is generated.
The pointer network model is pre-trained, which is essentially a neural network for solving a regression task, and after the network training is finished, a circuit board soldering path can be automatically generated only by one-time forward propagation operation.
Specifically, distance weights are determined according to whether the soldering points of the electronic equipment belong to the same confidence interval or not for different soldering points; calculating the actual distance between different tin soldering points by adopting a pre-trained pointer network model according to the distance weight of the electronic equipment; the actual distance between the different soldering points determines the circuit board soldering path.
The pointer network model is a network model which is specially aimed at the discrete state of output and has one-to-one correspondence relation with the input position, and the traditional pointer network model is used for solving the problem of PCB circuit board soldering point path planning by using supervised learning, namely an excellent algorithm is required to train an excellent route for the PCB circuit board soldering point path planning, and the performance upper limit is also the traditional algorithm. When the algorithm is used for training the parameters of the pointer network, the reinforcement learning Actor-Critic algorithm is adopted to optimize the pointer network, considering that the total distance of each calculation plan is a low-cost matter for a computer.
In this embodiment, given an input diagram, n welding points are represented as two-dimensional space s { (x) i ,y i ) S is expressed as the coordinate of a welding point in a two-dimensional space, i is the sequence of the point, and the actual distance S of each execution strategy pi is set as:
the component groups present in this space are each designated by A 1 ,A 2 ,A 3 .., the corresponding soldering point coordinate of each component set satisfies:
(x m ,y m ),(x m+1 ,y m+1 ),(x m+2 ,y m+2 )...∈A 1
in order to encourage the satisfaction of the prior welding strategy of the pins of the same component, the actual distance between two tin welding points of the same component set is multiplied by a preferential coefficient lambda (namely distance weight), and the size of the lambda depends on the priority of continuous welding with the pins of the component. So that the actual distance S π '：
The pointer network model used in the present embodiment is generated by training in advance. Specifically, a point set training set is generated by a program.
A C belonging to the area of the drag welding process is generated on a circuit board without any material (namely a blank picture) 0 Pin diagram of the component. For imitating C 0 And welding points of the element. As shown in FIG. 3, a pin c is randomly selected in a proper interval of the circuit board 0 Coordinate (x) of (2) 0 ,y 0 ) Of course, we also ensure that the generation of this coordinate does not cause the generated element to appearOutside this board. Component C 0 The width of the stitch is d w Element C 0 Length d between adjacent stitches l . These d w ，d l The parameters of the network are not particularly accurate, certain errors can improve the generalization performance of the network,
meanwhile, in order to enable the elements to be placed obliquely, the oblique angle theta is set, and a pin c can be calculated 7 Coordinate (x) of 7 ,y 7 ) Is (x) 0 -sinθdw,y 0 + cos θ dw), pin c 1 Coordinate (x) 1 ,y 1 ) Is (x) 0 +cosθdl,y 0 +sinθdl)。
But this is not the entirety of the pictures of one training set. In the program generation element C 0 Other components C of the same type can also be produced in the same way 1 ，C 2 ，C 3 For elements belonging to a spot welding area, the weld spot is generated directly in our training set with random coordinates within the interval. Of course, after generation, another simple formula is used to eliminate the random component C 0 ,C 1 ，C 2 .. points of internal balance, in C 0 Component as an example, the component pin c 0 ,c 3 ,c 4 ,c 7 Respectively is (x) 0 ,y 0 ),(x 3 ,y 3 ),(x 7 ,y 7 ) The original coordinate system is rotated counterclockwise by θ degrees. At this time, the direction corresponding to the long side of the original element becomes the x-axis in the new coordinate system.
x'＝x·cos(θ)+y·sin(θ)
y'＝y·cos(θ)-x·sin(θ)
C is to 0 ,c 3 ,c 7 The coordinates of the welding points are respectively (x) 0 ,y 0 ),(x 3 ,y 3 ),(x 7 ,y 7 ) Substituting the above formula for the cn point coordinate (x) n ,y n ) Satisfies the following conditions:
x′ 0 ≤x n ≤x′ 3 and y′ 0 ≤y n ≤y′ 7
then the point is removed from the training set, which is advantageous to speed up our training.
When the current position is at a tin soldering point, the time sequence is i, the pointer network available in the pointer network part can be output by softmax, pi i The strategy is the current strategy at the moment i, pi is the welding arrangement sequence of welding points of the whole circuit board, and the strategy pi function is formed by a lower pointer network.
The purpose of reinforcement learning is to update the strategy parameter theta of the strategy pi, and the purpose is to make the total planned path shortest after the strategy pi is executed in each path planning. And S (pi | S) represents the total distance of the welding spot path planning under the current strategy pi.
The expectation of the total distance S (S | S) is expressed by a performance index J (θ | S) containing a policy parameter θ:
since it is desirable to have a greater expectation that the total distance S (π | S) path will be shortest, J (θ | S) is first biased with respect to the policy parameter θ, which is defined according to the well-known formula of reinforcement learning (Williams, 1992):
p(π i |s i ) Is the probability that at i-pad location, the strategy pi selects the next pad, which is the learning goal of the pointer network, and p (pi | s) is all the probabilities when a given point s is set.
b(s) is a baseline function independent of pi, reduces the variance of the gradient by estimating the expected run length, is a gradient update parameter responsible for adjusting the gradient update rate. The algorithm enables the gradient of the performance index J (theta | s) to be increased by updating the strategy parameter theta, namely, the expectation that the total distance of the planned path is shortest is improved.
The pointer network is based on an attention mechanism which is divided into an encoding end and a decoding end.
Where ej is the hidden layer output of the encoder in time sequence j, d i Is the decoder output in the hidden layer of time sequence j, where W 1 ,W 2 ,v T Are parameters of fixed dimension that make us need to be trained. For u is paired j Performing softmax can obtain the probability vector output to the corresponding welding point, and the size of the output vector should be consistent with the size of the input vector of the welding point coordinate set given by us.
The encoding end firstly converts the soldering point coordinates into discrete vectors (embedding) and then inputs the discrete vectors into a hidden layer of an encoder. The output of this step is weighted a, either scalar or vector, each time a weld point is passed, and each weighted product is finally summed. The decoder part also has the same hidden layer. At each time step i, a pointer mechanism is used to generate the probability distribution of the next visit to the solder joint.
Because the pointer network model algorithm is essentially a neural network fitting process, once the model training is finished, the pointer network model algorithm is equal to a classifier which finds the shortest welding path on a circuit board, and the pointer network model algorithm does not need to be learned again when meeting different circuit boards and does not need to have an optimization process like the traditional optimization algorithm, so the generalization is better, and the pointer network model algorithm can be suitable for different circuit boards.
In particular, the present embodiment is able to handle the structure of any problem sampled from a given distribution well, rather than training a separate model for each instance. This means that even if a new circuit board is changed, which has a similar number and location of pads as we generated in the training set, then the trained strategy will work well and we can solve the problem immediately without having to retrain each circuit board.
Optionally, because a large number of circuit board images need to be acquired as sample images for deep learning training, and the acquisition cost of the circuit board images is greatly increased by acquiring a large number of circuit board images, in order to further reduce the acquisition cost of the circuit board images, a training set with a large number of soldering points is generated by referring to a real circuit board, so that the training efficiency of deep learning is greatly increased on the premise of ensuring that the actual model of the circuit board is met.
Specifically, when the simulated circuit board image is generated by referring to the real circuit board, the simulated circuit board image is generated according to the position of the simulated soldering point of the electronic equipment, and the simulated circuit board image of the electronic equipment is determined as the sample image.
Example two:
fig. 4 is a block diagram of a path planning apparatus to which a circuit board is soldered according to the second embodiment. The device can execute all or part of the steps of any one of the circuit board soldering path planning methods. The device comprises:
and the soldering point extraction module 1 is used for extracting the position coordinates of the soldering points from the circuit board image of the electronic equipment to obtain a soldering point set according to the numerical values of all the pixel points in the circuit board image.
And the confidence interval division module 2 is used for identifying the attribution elements of the tin soldering points in the tin soldering point set on the circuit board and dividing the confidence intervals in the tin soldering point set.
And the soldering path generation module 3 is used for adopting a pre-trained pointer network model to forward propagate the circuit board soldering point sets for dividing the confidence intervals to generate a circuit board soldering path.
Example three:
the third embodiment of the present application provides an electronic device, which may be a mobile phone, a computer, a server, or the like, and may perform all or part of the steps of any one of the above-described circuit board soldering path planning methods. The electronic device includes:
a processor; and
a memory communicatively coupled to the processor; wherein the content of the first and second substances,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method according to any one of the above exemplary embodiments, which will not be described in detail herein.
In this embodiment, a storage medium is also provided, which is a computer-readable storage medium, such as a transitory and non-transitory computer-readable storage medium including instructions. The storage medium includes, for example, a memory of instructions executable by a processor of an electronic device to perform a method of path planning for soldering of a circuit board as described above.
The above description is only a preferred embodiment of the present application and should not be taken as limiting the present application, and any modifications, equivalents, improvements, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.
Claims (9)
1. A method for path planning for circuit board soldering, the method comprising:
training a pointer network model by using a pointer network as a path planning backbone network;
and planning paths of the welding point set on the soldering circuit board by adopting the trained pointer network model to generate a soldering path of the circuit board.
2. The method of claim 1, wherein when the pointer network is used as a path planning backbone network for training a pointer network model, a reinforcement learning algorithm Actor-Critic is used to perform gradient descent on the pointer network of the backbone network, and during the descent process, a loss function of the Critic network is modified by using the following weighted distance formula:
3. the method of claim 2, wherein the pointer network model is trained using the pointer network as a path planning backbone network, the method further comprising:
acquiring sample images of more than ten thousand levels;
adding a component pin point set into the sample image;
dividing confidence intervals in the sample image according to the component pin point set;
and calculating the distance between the point sets by adopting the weighted distance summation formula in the confidence space.
4. The method of claim 1, wherein when using the pointer network model to route a set of solder joint points on a soldered circuit board,
extracting a welding spot coordinate point set on the soldering circuit board;
and taking the welding point coordinate point set as input, and carrying out soldering path planning by adopting the network parameter forward propagation in the pointer network model after training.
5. The method of claim 4, wherein the step of extracting a set of solder joint coordinate points for the soldered circuit board comprises:
extracting the numerical value of a pixel point in the circuit board image;
extracting a central coordinate of which the pixel value is approximate to the pixel value of a real welding spot from the circuit board image;
and determining the center coordinates as corresponding welding point coordinates.
6. The method of claim 1, wherein, in generating training samples:
generating part of simulated soldering points according to the layout of pins of different elements;
randomly generating partial soldering points;
and generating a training sample according to the soldering points.
7. A circuit board soldering path planning apparatus, the apparatus comprising:
the tin soldering point coordinate extraction module is used for extracting the coordinates of the center positions of the tin soldering points from the circuit board image according to the difference of pixels in the circuit board image to obtain a tin soldering point set;
and the soldering path generation module is used for adopting a pre-trained pointer network model to forward propagate the soldering point set to generate a circuit board soldering path.
8. An electronic device, characterized in that the electronic device comprises:
a processor; and a memory communicatively coupled to the processor; wherein the memory stores readable instructions which, when executed by the processor, implement the method of any one of claims 1-6.
9. A computer-readable storage medium, on which a computer program is stored which, when executed, implements the method of any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210725876.7A CN114980527B (en) | 2022-06-24 | 2022-06-24 | Circuit board soldering path planning method and device and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210725876.7A CN114980527B (en) | 2022-06-24 | 2022-06-24 | Circuit board soldering path planning method and device and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114980527A true CN114980527A (en) | 2022-08-30 |
| CN114980527B CN114980527B (en) | 2024-02-09 |
Family
ID=82965682
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210725876.7A Active CN114980527B (en) | 2022-06-24 | 2022-06-24 | Circuit board soldering path planning method and device and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114980527B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116423002A (en) * | 2023-06-13 | 2023-07-14 | 苏州松德激光科技有限公司 | Automatic correction method and system for welding position |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019056749A1 (en) * | 2017-09-19 | 2019-03-28 | 北京嘉楠捷思信息技术有限公司 | Printed circuit board structure, and wiring method therefor |
| CN110610199A (en) * | 2019-08-22 | 2019-12-24 | 南京理工大学 | Automatic optical detection method for printed circuit board resistance element welding spots based on svm and xgboost |
| US20210097449A1 (en) * | 2020-12-11 | 2021-04-01 | Intel Corporation | Memory-efficient system for decision tree machine learning |
| CN112935442A (en) * | 2021-02-03 | 2021-06-11 | 中国科学院深圳先进技术研究院 | Method, device, equipment and medium for determining parameters of automatic welding process |
-
2022
- 2022-06-24 CN CN202210725876.7A patent/CN114980527B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019056749A1 (en) * | 2017-09-19 | 2019-03-28 | 北京嘉楠捷思信息技术有限公司 | Printed circuit board structure, and wiring method therefor |
| CN110610199A (en) * | 2019-08-22 | 2019-12-24 | 南京理工大学 | Automatic optical detection method for printed circuit board resistance element welding spots based on svm and xgboost |
| US20210097449A1 (en) * | 2020-12-11 | 2021-04-01 | Intel Corporation | Memory-efficient system for decision tree machine learning |
| CN112935442A (en) * | 2021-02-03 | 2021-06-11 | 中国科学院深圳先进技术研究院 | Method, device, equipment and medium for determining parameters of automatic welding process |
Non-Patent Citations (1)
| Title |
|---|
| 陈晨;王亚立;乔宇;: "任务相关的图像小样本深度学习分类方法研究", 集成技术, no. 03 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116423002A (en) * | 2023-06-13 | 2023-07-14 | 苏州松德激光科技有限公司 | Automatic correction method and system for welding position |
| CN116423002B (en) * | 2023-06-13 | 2023-10-31 | 苏州松德激光科技有限公司 | Automatic correction method and system for welding position |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114980527B (en) | 2024-02-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106650630B (en) | A kind of method for tracking target and electronic equipment | |
| CN108304835A (en) | character detecting method and device | |
| CN108898086A (en) | Method of video image processing and device, computer-readable medium and electronic equipment | |
| CN110059750A (en) | House type shape recognition process, device and equipment | |
| CN110674714A (en) | Human face and human face key point joint detection method based on transfer learning | |
| CN108898185A (en) | Method and apparatus for generating image recognition model | |
| CN110636715A (en) | Self-learning-based automatic welding and defect detection method | |
| CN108229479A (en) | The training method and device of semantic segmentation model, electronic equipment, storage medium | |
| CN108197623A (en) | For detecting the method and apparatus of target | |
| CN112529178A (en) | Knowledge distillation method and system suitable for detection model without preselection frame | |
| CN106845487A (en) | A kind of licence plate recognition method end to end | |
| CN109934115A (en) | Construction method, face identification method and the electronic equipment of human face recognition model | |
| CN108334805A (en) | The method and apparatus for detecting file reading sequences | |
| CN109214471A (en) | Evaluate the method and system of the written word in copybook of practising handwriting | |
| CN110689000B (en) | Vehicle license plate recognition method based on license plate sample generated in complex environment | |
| CN107622252A (en) | information generating method and device | |
| CN110633843B (en) | Park inspection method, device, equipment and storage medium | |
| CN113939791A (en) | Image labeling method, device, equipment and medium | |
| Barman et al. | Shape: A novel graph theoretic algorithm for making consensus-based decisions in person re-identification systems | |
| CN109657792A (en) | Construct the method, apparatus and computer-readable medium of neural network | |
| CN114980527A (en) | Circuit board soldering path planning method and device and electronic equipment | |
| WO2023116565A1 (en) | Method for intelligently designing network security architecture diagram | |
| CN109389105A (en) | A kind of iris detection and viewpoint classification method based on multitask | |
| CN114777792A (en) | Path planning method and device, computer readable medium and electronic equipment | |
| CN107369086A (en) | A kind of identity card stamp system and method |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |