Precision PCB Fabrication, High-Frequency PCB, High-Speed PCB, Standard PCB, Multilayer PCB and PCB Assembly.
The most reliable PCB & PCBA custom service factory.
PCBA Tech

PCBA Tech - SMT solder joint system and solder paste for surface assembly

PCBA Tech

PCBA Tech - SMT solder joint system and solder paste for surface assembly

SMT solder joint system and solder paste for surface assembly

2021-11-07
View:376
Author:Downs

1. Work flow of SMT solder joint system

The working process of the system can be expressed as follows: (1) First, according to the input solder joint type, apply the corresponding training sample to train the neural network, and save the connection weight matrix as knowledge in the knowledge base

The working process of the system can be expressed as follows:

(1) First, according to the input solder joint type, apply the corresponding training samples to train the neural network, and save the connection weight matrix as knowledge in the knowledge base.

(2) Read in the deviation value between actual solder joint shape parameters and reasonable solder joint shape parameters, and enter the database for query. If the query is successful, the corresponding control strategy will be given to further adjust the process parameters or eliminate the fault. If the query is unsuccessful, Then use the knowledge in the knowledge base to give the operation result through the inference algorithm.

(3) Perform test verification and evaluation on the recommended values given by the system. If the results are satisfactory, store them in the database, and perform solder joint assembly quality control based on the results, otherwise return to repeat the above process.

pcb board

(1) Choice of network model

In the SMT solder joint quality analysis and evaluation expert system based on neural network, the knowledge base is used to store the connection weights between each neuron, which is represented by a digital weight matrix. When constructing the knowledge base, a very important task is the choice of the network model, that is, determining the expression mode and learning algorithm of the neural network. In the learning module, I hope to find the corresponding relationship between the deviation value of the actual solder joint and the reasonable solder joint shape parameter of the SMT product and the control amount that should be adopted. The input and output parameters of the system can be continuously changed, and the control conclusions obtained under certain input are required in the training, and the error with the conclusions given by SMT process experts is small enough, so teachers are required to learn. Based on the above requirements, the error back propagation network (ie BP network) is selected here to achieve,

(2) BP algorithm

The BP algorithm corresponding to the BP network is composed of two parts: the forward transmission of information and the backward transmission of errors. The following uses a two-layer BP network as an example to illustrate the principle of the BP algorithm. Suppose the input of the network is force, and the number of neurons in the input layer, hidden layer and output layer are r, si, and S2 respectively; the weight coefficient from the input layer to the hidden layer is called the transfer function of gangrene and the hidden layer to The weight coefficient of the output layer is 2, 2, and the transfer function is 7; the output of the network is A, and the target vector is T; bi and b2 represent the input layer and hidden layer neuron threshold values, respectively. The BP algorithm process can be summarized as follows:

1. Initialize the network connection weight and assign it to a small initial value;

2. Provide input and expected output;

3. Calculate the actual output, including the nodes of the output layer and the hidden layer;

4. Adjust the weights, use the regression algorithm, first start from the output layer, and then return to the hidden layer until the first hidden layer. If the error termination condition is met, the adjustment is stopped.

(1) Solder joint shape comparison model

The SMT solder joint quality analysis and evaluation expert system based on neural network uses the deviation between the actual solder joint shape and the reasonable solder joint shape as the basis for judging the quality of the solder joint. This deviation can be obtained by comparing the parameters of the solder joint shape. There are many methods for comparing solder joint shape parameters. The following is one of the simpler and direct comparison methods. For a particular SMT solder joint, its morphological parameters can be arranged in a certain order, expressed as a matrix of NX1 (N represents the number of solder joint morphological parameters).

Second, the requirements of SMT surface assembly for solder paste

1) The solder paste should have good storage stability. After the solder paste is prepared, it should be stored at room temperature or refrigerated conditions for 3 to 6 months before printing without changing its performance. (2) What the solder paste should have during printing and before reflow heating

1) The solder paste should have good storage stability. After the solder paste is prepared, it should be stored at room temperature or refrigerated conditions for 3 to 6 months before printing without changing its performance.

(2) The properties that the solder paste should have during printing and before reflow heating:

1. The solder paste should have excellent mold release during printing.

2. The solder paste is not easy to collapse during and after printing.

3. The solder paste should have a certain viscosity.

LED high power light board patch 3

(3) The properties of solder paste during reflow heating:

1. Should have good wetting performance.

2. No or minimum solder balls are formed.

3. There should be less solder spatter.

(4) The properties that the solder paste should have after SMT reflow soldering:,

1. The lower the solid content in the flux, the better, and it is easy to clean after welding.

2. High welding strength.