Issue link: https://iconnect007.uberflip.com/i/1529411
NOVEMBER 2024 I PCB007 MAGAZINE 25 parameters: n number of trials and p probabil- ity of "passing" with 0 ≤ p ≤ 1 7, 8 . Statistical Tools for Analyzing and Monitoring Defects and Defectives We use analyses based on a Poisson proba- bility model to evaluate defects. ese analyses evaluate the rate of defects in a process. Stan- dard statistical tools include the C Chart, used to monitor the number of defects where each unit can have multiple defects. You should use a C Chart only when your subgroups are the same size. U Chart is used to monitor the number of defects per unit, where each unit can have multi- ple defects and various subgroup sizes. U Chart Diagnostic is used to test for overdispersion or underdispersion in your defects data. (Note that overdispersion and underdispersion are mutually exclusive.) Overdispersion can cause a traditional U Chart to show increased points outside the control limits. Underdispersion can cause a conventional U Chart to show too few points outside the control limits. e Laney U' chart adjusts for these conditions. Poisson Capability Analysis determines whether the defects per unit (DPU) rate meets customer requirements 9 . We use analyses based on a binomial probabil- ity model to evaluate defectives. ese anal- yses evaluate the pro- portion of defectives in a process. Standard sta- tistical tools include the P Chart, used to mon- itor the proportion of defective units where each unit can be classi- fied into one of two cat- egories: pass or fail, and with various subgroup sizes. P Chart Diag- nostic is used to test for overdispersion or underdispersion in your defectives data; note that overdispersion and underdispersion are mutually exclusive. Over- dispersion can cause a traditional P chart to show increased points outside the control lim- its. Underdispersion can cause a conventional P chart to show too few points outside the control limits. e Laney P' chart adjusts for these con- ditions. NP Chart is used to monitor the num- ber of defective units; each unit can be classified into one of two categories: pass or fail, and has various subgroup sizes. Note: if your subgroup sizes are not equal, the center line will not be straight. Binomial capability analysis is used to determine whether the percentage of defective units meets customer requirements 9 . A Worked Example A young engineer understands the impor- tance of yield statistics concerning customer sat- isfaction and continuous process improvement. Unfortunately, they have no data on defects. Still, they do have limited data on defective units processed through the I/L, making holes conductive (MHC), and O/L processes. ey review the last eight weeks of production data as a starting point and calculate the FTY and RTY statistics (Table 1). Because we are work- Table 1: Eight weeks of production data