Their research proved that the cost function was convex with respect to the concurrent (or overlap) degree between design tasks and that it must have a minimum value at a unique price Ruxolitinib optimum point. Huang and Gu [13, 14] viewed the product development process as a dynamic system with feedback on the basis of feedback control theory. The dynamic model and its design structure matrix were developed. The model and its design structure matrix could be divided farther to reflect the interaction and feedback of design information. The

mode and direction of the development process could be selected to satisfy constraints of process data flow and process control. A fuzzy evaluation method was presented to evaluate the performance of the dynamic development process; this allowed the development process to be optimized based on reorganizing design constraints, reorganizing design processes, and reorganizing designer’s preferences.

Finally, an application shows that modeling the product development process as a dynamic system with feedback was a very effective method for realizing life cycle design, optimizing the whole development process, improving the degree of concurrent, speeding information flow, and reducing modification frequency. However, due to complexity of product development, this model did not consider the currency and overlapping among tasks. Its efficiency needs further study and verification. Zhang et al. [15] constructed a new method to measure the coupled strength and to calculate the first iteration’s gross workload of a different sequence of coupled tasks, thereby ascertaining the best sequence of coupled tasks based on existent research. However, this model may not correspond to real-world product development process and it is also dependent on expert’s experiences. Moreover, Xiao et al. [16] adopted analytic hierarchy process (AHP) to deal with coupling tasks, which might cause quality

loss. Smith and Eppinger [17, 18] set up two different iteration models based on DSM. One is the sequential iteration model and the other is the parallel iteration model. The Drug_discovery former supposed that coupled tasks were executed one after the other and rework was governed by a probabilistic rule. Repetition probabilities and task durations were assumed constant in time. The process was modeled as a Markov chain and the analysis could be used to compute lead time for purely sequential case and to identify an optimal sequence of the coupled tasks to minimize iteration time. The main limitation of this model is that how to determine repetition and rework probabilities is difficult. The latter supposed that the coupled design tasks were all executed in parallel and iteration was governed by a linear rework rule. This model used extended DSM called work transformation matrix (WTM) to identify the iteration drivers and the nature and rate of convergence of the process. WTM has been popularly used in many areas.