2. In DFSS, the internal requirements of enterprises are called internal customer requirements, and some requirements are very important indicators, such as cost. Tianxingjian Consulting requires enterprises to incorporate QFD, decompose it layer by layer, and finally refine it to each part.
3.QFD is subdivided level by level, so if the customer provides system indicators, you can certainly start here.
3.FMEA is a cumulative process. A small product of some enterprises has thousands of FMEA lines and hundreds of failure modes. This is not to say that their products have many problems (because only a few of them have high RPN values), but that they consider the possible problems of their products very comprehensively, so the probability of uncontrollable problems in the market is greatly reduced. At present, the comprehensiveness of FMEA in many enterprises needs to be accumulated and strengthened.
4. According to the "28" principle, determine how many measurement systems need to be analyzed, and 20% of the indicators of the system are the key parameters of the system. All the CTQ obtained by QFD decomposition are sorted from big to small, and the previous 20% CTQ is selected as the key CTQ of measurement system analysis.
5. results about DOE. The goal of DOE is to find the relationship between factors and indicators, so the results of various combinations of DOE may exceed the standard or fall below the standard, because the purpose is to find the best situation under what combination. DOE does not do design acceptance, but explores the corresponding relationship. At present, if all the tests exceed the standard, it means that the optimal solution cannot be found within the current range. Need to adjust the factor level and reposition the search area.
6. Of course, the search of transfer function can also be obtained according to other similar models, but if there are obvious differences in structure, it cannot be recursive.
7.① The percentage of R-Sq in regression analysis is not high, which is related to the wrong factor screening method. After removing the interactive items, the R-sq (adjustment) reached more than 97%.
② All the factors are irrelevant, which is related to the wrong factor screening method. After removing the interactive items, the correlation has already appeared. In addition, if the related items can't appear in the end, it shows that the difference between the combinations of DOEs is natural error and not significant, so the factors in DOEs are not the key factors.
(3) It is unclear how to analyze the obtained test results. After obtaining the test results, please find the partial derivative of the related factors according to the constructed regression equation, and then find the variance range of Y. ..
(4) When eliminating cross effects, the principle of deletion is to keep the single factor at the minimum value of standardized effects from top to bottom in the pareto chart of standardized effects, and delete all other interactions. Is it okay to do this?
The deletion principle should be:
The deletion order is "bottom-up" according to pareto chart, that is, starting from P value >; 0.05 delete P & gt0.05 at most, and delete the interactive items first. When the main effect (single factor) has interaction, delete one at a time according to the principle of non-deletion, and recalculate after deletion. Don't delete too much at once.
8. When calculating the deviation of CTQ, use the method of partial derivative to find out whether the values of the three factors are fixed values or not, and whether they should be treated as constants. The parameters with fixed values can be used as constants in the model, not as variables, so there is no need to take partial derivatives.
9. Theoretically, whether the index part with smaller Z value after partial derivative transfer can be used as the direction of optimization is like this. Of course, when determining the improvement goal, we should consider not only the existing level (shortcomings or bottlenecks), but also the improvement cost and the size of the improvement space. It is necessary to comprehensively consider from multiple dimensions and choose the best improvement direction.