Statistical Process Control in Electronic Manufacturing

To control the quality of product, we implement SPC (Statistic Process Control) to ensure manufacturing process is performed in controlled procedure. Even though, there will be no perfect mass product even if our quality control is very stringent. We have to balance between price and manufacturing cost since quality is meeting customer expectation. The more stringent quality control, the more cost required

We use three tools to control manufacturing process :
- p-Chart
- u-Chart
- Pareto Analysis

p-Chart :

Is use to monitor the proportion of nonconforming units in a sample : conform or non-conform, yes or no, etc. Nonconformity is defects or occurrences found in product set. It can be described as any characteristic that is present but should not be, or any characteristic that is not present but should be. I will use our p-Chart for discussion purpose.

We use FW (fiscal week) for our production schedule. Week 5.05 is 5th week and day 5 (Friday). In 2009, FW5.05 is Friday, 30 January 2009. At FW5.05, we made 3988 sets remote control (our main product). At that day, we got 91 sets reject (Non Conformance – NC) or the proportion = 2.29%. We plot this proportion percentage on chart. At FW8.04 we made 5994 sets with 183 sets reject or 3.05%. We also plot this number into chart.

From above table / chart, we calculate UCL (Upper Control Limit) and LCL (Lower CL). These two lines are used to determine whether a manufacturing process is under control or not. If the chart is under control then it can be used to predict the future performance of the process. If the chart is not in control, the pattern it reveals can help determine the source of variation to be eliminated to bring the process back into control. Ideally UCL and LCL are taken from 1 month observation, but it can be made faster (i.e. 2 weeks). The longer the observation, the better UCL/LCL can be got.

Based on observation started from FW5.05 until FW15.01, we determine UCL= 3.16 (red dash) and LCL.= 1.87 (black dash). These control parameter are used to control next month process. In average, reject deviation from FW15.02 – FW19.04 is considered normal, only at 15.02 and 16.01 reject sets quantity are good.

If we can observe another observation period (i.e. another 30 days observation), we may get different UCL and LCL.

u Chart :

To control the number of non conformity per unit, or we say defect per unit (DPU), we use u-Chart. In our manufacturing process, the populated PCB (Printed Circuit Board) will be dipped into solder pot. After this process, there will be many defects (nonconformities), i.e. short soldered, not soldered, etc. Each defect will be counted. Let’s say there are 10 pcs PCB. There is 5 defects / nonconformities in PCB no. 1 and  3 defect in PCB no. 2 but no defect found on the rest PCB (8 pcs). We say that in 10 pcs PCB, there is 8 defects, or defect per unit = 0.8. Beside DPU, we also record defect classification and number of occurrences and then analyze the defects pattern. By knowing defect pattern, we are able to determine if the process is stable and predictable, as well as to monitor the effects of process improvement.

Please visit http://syque.com/quality_tools/toolbook/Control/do_u_calc.htm to learn how to calculate UCL and LCL. By using this formulas, we get UCL = 1.61 and LCL = -0.06

Pareto Analysis :
Another tool can be used to control manufacturing process is pareto analysis or 80 : 20 analysis. Theoretically, if we can solve top 20% reject, its mean we can solve 80% problem. In practice, 80 : 20 does not mean 80% and 20%, but if we can capture the most two or three problems, we can solve most of the problem

Beside reject percentage, we also have to record reject cause. Based on our reject cause of above table / chart, we can group into 6 reject categories. Please see below picture

You can see there are 3 most problems, no. 1, 2 and 4. For the first time we have to concentrate to solve most 2 – 3 rejects. After these reject can be solved, then we go to next reject (maybe number 5 and 3)