|
Robust multi-objective economic-statistical design of nonlinear profile monitoring under parameter uncertainty and multiple assignable causes: An MOGWO-based approach
|
| |
|
|
|
چکیده: (2 مشاهده) |
Monitoring nonlinear quality profiles becomes difficult when profile coefficients are correlated, several assignable causes may occur, and process and cost parameters are uncertain. This study develops a robust multi-objective economic-statistical design that integrates a third-order polynomial profile, an EWMA-Hotelling T² monitoring scheme, cause-specific competing risks, scenario-based parameter uncertainty, and a simulation-based regenerative cost model. The decision variables are sample size, sampling interval, control-limit multiplier, and EWMA smoothing parameter. Monte Carlo simulation is used to estimate in-control and out-of-control run-length behavior and calendar-time detection delay, while a constrained mixed-variable multi-objective grey wolf optimizer constructs feasible Pareto alternatives. The proposed search procedure is evaluated descriptively against NSGA-II and MOPSO under a matched verification setting. The results indicate that MOGWO provides a competitive balance between Pareto-front convergence and coverage, although competing algorithms retain advantages on individual diversity measures. Policy-level and sensitivity analyses further show that detection performance and economic cost are strongly influenced by process dispersion, shift magnitude, and out-of-control losses. The proposed framework provides practitioners with a transparent set of monitoring policies for balancing cost, false-alarm protection, detection speed, and robustness under uncertain operating conditions.
|
|
| |
|
|
متن کامل [PDF 1135 kb]
(8 دریافت)
|
نوع مطالعه: پژوهشي |
موضوع مقاله:
عمومى
دریافت: 1405/5/27 | پذیرش: 1405/7/10 | انتشار: 1405/7/26
|
|
|
|
|
|
|
| ارسال نظر درباره این مقاله |
|
|
|
