Reflow Process Parameters Analysis and Reliability Prediction Considering Multiple Characteristic Values

As a major step surface mount technology, reflow process is the key factor affecting the quality of the final product. The setting parameters and characteristic value of temperature curve shows a nonlinear relationship. So parameter impacts on characteristic values are analyzed and the parameters adjustment process based on orthogonal experiment is proposed in the paper. First, setting parameters are determined and the orthogonal test is designed according to production conditions. Then each characteristic value for temperature profile is calculated. Further, multi-index orthogonal experiment is analyzed for acquiring the setting parameters which impacts the PCBA product quality greater. Finally, reliability prediction is carried out considering the main influencing parameters for providing a theoretical basis of parameters adjustment and product quality evaluation in engineering process.


Introduction
In the era of rapid development for communication and information technology, electronic production has gradually become an indispensable necessity in daily life.To quickly adapt to market demand, electronic product design and manufacturing technology advances rapidly.Surface mount technology (SMT) has become the main way of electronic product manufacturing for its good coupling performance and high production efficiency.Industrial production usually aims at meeting customer requirements for product quality.As the main process in SMT, reflow profile quality reflects the final product quality directly in reflow soldering process [1].During the reflow process, it is difficult for reflow profile to fully meet the customer's requirements.For this reason, we need to control the curve error within a reasonable range.Currently, the main method for reflow profile prediction is try and error, which requires that the operator has a wealth of experience [2].Since the reflow process has a certain randomness which is affected by various uncertainties and the environment, it is necessary for reflow profile setting and analysis from the perspective of reliability.Considering the complexity of reflow process [3,4], the paper conducts the study with the production examples from a company in Kunshan.
In order to investigate the influence on characteristic values, orthogonal design for parameters is carried out first to determine the main setting parameters and its influence.The study provides the basis for the reflow process parameter adjustment.Then the evaluation index for reflow profile characteristic values is defined to fit the reliability formula with Weibull distribution.The reliability of reflow profile characteristic values is also evaluated.Consequently, reliability analysis of reflow profile provides a theoretical basis for determining the characteristic value error range.

Reflow parameters analysis based on orthogonal experiment 2.1 Orthogonal Experimental Design
Orthogonal Experimental design is a design approach which arranges and analyzes multivariate experiment using normalized orthogonal [5].Study of the reflow parameters impact on reflow profile and key indicators based on orthogonal experiment is able to determine the greatest impact parameter on all key indicators and their impact direction [6].
For the reflow process, there are many parameters to characterize the quality of reflow profile.Besides, each input parameter is the factors affecting the final quality of the reflow profile comprehensively.Therefore, the use of orthogonal experiment method can efficiently analyze the reflow process with multiple indicators.

Orthogonal Experimental Design
From the analysis, the setting parameters of the reflow process are the key factors to affect the reflow profile.So the setting parameters of eight oven zones(( )) and the belt speed(m/min)marked A~J are selected as input parameters.Three levels are determined for each factor in the experiment.The characteristic values of reflow profile are peak temperature ( ) soaking time(s), climbing time(s), melting time(s), cooling rate(m/min) and infiltration temperature ( ) marked K~P.The factors and their levels are listed in table 1.  2.

Table 2. Orthogonal design& experimental results
No K ij represents the mean value of experimental result for the jth column and the ith level.T ij is the sum of experimental results for the jth column and the ith levels.K ij (max) is the maximum K ij value in the jth column.K ij(min) is the minimum K ij value in the jth column.c is range value of the jth column.
Range analysis is capable of studying the degree of setting parameters influence on the characteristic values.Thus, range analysis is implemented on orthogonal test results in table 2. K value and R value of each characteristic value is computed by formula (1) and formula (2).According to range analysis, the main factors affecting the peak temperature is temperature of zone 7, temperature of zone 6, belt speed and temperature of zone 4 in succession; the main factors affecting the soaking time is belt speed, temperature of zone 6, temperature of zone 3, and temperature of zone 4 in succession; the main factors affecting the climbing time is temperature of zone 6, belt speed, temperature of zone 5 and temperature of zone 4 in succession; the main factors affecting the melting time is belt speed, temperature of zone 6, temperature of zone 8, and temperature of zone 7 in succession;the main factors affecting the cooling rate is temperature of belt speed, zone 8, temperature of zone 7 and temperature of zone 2 in succession; the main factors affecting the infiltration temperature is temperature of zone 6, temperature of zone 5, temperature of zone 7 and temperature of zone 8 in succession.Analysis of K value is shown in figure 1.
It is known from figure 1 that the output distribution shows a range of differences with different factor influence degree.The factor with a relatively large difference is corresponding to the important one, which is consistent with the characteristic values in the oven zone accordingly during the reflow process.From this point of view, we conclude that the analyzed result meets the actual production process.To sum up, belt speed, zone 7 and zone 8 are the major factors to influence the reflow profile, which needs to monitor particularly in the production process.During the production process, the above findings may provide a theoretical reference for the production of parameter adjustment.

Reliability of characteristic value index
The sample No. 2Q012D000-600-GB/T-A-HF is used in the experiment for analysis.The setting parameters are listed in table 3.In the experiment, four major parameters are used as random variables for reliability analysis.From the above analysis, the major parameters are temperature of zone 4, zone 6, zone 7 and belt speed.Due to differences in the production conditions, errors are produced for various parameters inevitably, which affects product reliability significantly.For analyzing the randomness of parameters, the above major parameters are supposed as random variables which obey normal distribution as table 4. The probability level is chose as p 1 = 0.01 p 2 = 0.5 p 3 = 0.99.Then the sample value is calculated according to Box-Behnken matrix sampling method [7].Furthermore, characteristic value is calculated with BP-Neural Network.For evaluating the fitting reflow profile comprehensively, reflow profile quality evaluation index is defined as equation ( 4).According to the definition of evaluation index, it is known that the smaller f(x) represents the product meeting customer requirements better.The calculated results are listed in table 5. xi represents the reliability of evaluation index.
1 D D D , the hypothesis H 0 is acceptable.Therefore, the reliability of evaluation index obeys the Weibull distribution, which is shown as follows.
2.256 exp exp 0.045 The fitting curve of reliability is shown in figure 2. It is obvious that the smaller evaluation index is, the higher the reliability is.
The fitting equation of reflow profile quality evaluation index reliability has significant practical value.On one hand, when the evaluation index is calculated according the characteristic value of the reflow profile, the reliability of the evaluation index can be calculated.Furthermore reliability of product can be acquired quantitatively as well.On the other hand, precision condition of characteristic value acquired by custom can also be determined with the index reliability analysis.(2) Reflow profile quality evaluation index is defined to calculated the characteristic value.Weibull distribution is applied for fitting the equation of index reliability.The reliability equation is capable of acquiring the product reliability with required evaluation index.

Figure 1 .
Figure 1.Analysis of K-value

Figure 2 .
Figure 2. Fitting curve of reliability

Table 1 .
Factor levels 27(3 13) orthogonal table is used for arranging 27 sets orthogonal experiments.Then the experiments are carried out in an electronics production company in Kunshan.The experimental results are recorded in table Range analysis aims at determining progression of factors based on range value calculated with mathematical statistics calculation method.There are m factors in the experiment according to orthogonal table.Experimental results are recorded by x1, x 2 ,…, x N .Each factor has t levels, and n experiments are carried out for each level.Therefore, the total experiment time N is calculate by N=tn.It is known that

Table 3 .
Temperature setting of reflow soldering oven

Table 4 .
Mean and standard deviation of all variation

Table 5 .
Sample points of Box-Behnken method and data of reliability tests S test is applied for testing the fitness of parameters estimation.The observed value level is taken as α=0.10, so D n,a =0.2377.Since it is known that