National Natural Science Foundation of China (61370058, 61170087)
In large system production line configuration, manual configuration is inevitable and hence easy to introduce nonconformities where configuration data inputted by configuration engineers violate predefined constraints(also known as conformance constraints). For large system production lines, such as cyber physical system(CPS) product lines, there are usually hundreds and thousands of configurable parameters, hundreds of conformance constraints, and complicated dependencies among the conformance constraints. Thus it is very challenging to resolve nonconformities in an efficient manner. As a first step to address this challenge, an automated nonconformity resolving recommendation approach(Zen-Fix) was presented in the previous work by this research, which relies on multi-objective search and constraint solving techniques. To further improve the search efficiency in such interactive CPS configuration process, this paper proposes a novel algorithm called DeIBEA, which combines differential evolution with IBEA(indicator-based evolutionary algorithm), and distinguishes feasible solutions from infeasible ones, generating offspring through the differential operation. Integrating Zen-Fix with DeIBEA can recommend nonconformity-free yet optimal solutions to configuration engineers. The cost effectiveness of DeIBEA(in the context of Zen-Fix) is empirically evaluated with a real-world case study, in which a configuration process is simulated containing 10189 search problems. Results show that:(1) Zen-Fix with DeIBEA can provide nonconformity resolving recommendation automatically in a quite efficient way;(2) Compared with IBEA, DeIBEA performs significantly better in terms of both time performance and search performance.