Đang chuẩn bị liên kết để tải về tài liệu:
Quality Management and Six Sigma Part 8

Tham khảo tài liệu 'quality management and six sigma part 8', kỹ thuật - công nghệ, cơ khí - chế tạo máy phục vụ nhu cầu học tập, nghiên cứu và làm việc hiệu quả | Design for Six Sigma DfSS in Software 133 larger sample say 100 devices then the product starts failing. So OCC value could be different based on the sample taken and has to be accordingly adapted when validating the results 8 From software development life-cycle perspective the DET value can take on different values for the same detection levels. For e.g. a control mechanism may have a high chance of detecting a failure mode making the DET value 4 as per the guideline. However based on whether that detection can happen in design itself or testing may vary the value. The team might give a higher vale for DET for something that can be detected only in testing as against that which can be detected in design. 4.3 Use of Statistics in software Often this is one of most important challenge when it comes to using concepts like DfSS for software. Many software requirements fall into the Yes No Pass Fail category so limit setting is fuzzy. Most of them would become critical factors CFs and not CTQs in the continuous data sense Predicting DPMO defects per million opportunities may be misleading out of limits . This is because the specifications limits in cases like responsiveness are soft targets. Just because it takes 0.5 seconds more than Upper Specification Limit to start-up does not necessarily classify it as a defective product. In Six sigma terms anything beyond Upper spec limit and less than Lower spec limit becomes a defect Random failures due to only software are rare due to which concept like Mean-Time-Between-Failures MTBF for software alone is questionable however it makes sense at overall product level No concept of samples - the same piece of code is corrected and used so advanced statistical concepts have to be applied with discretion However this does not mean that statistical concepts cannot be applied at all. The starting point is to challenge each specification to ensure if some numbers can be associated with it. Even abstract elements such as Usability .