A configuration-specific analytical evaluator for energy-aware capacity reconfiguration in two-class CONWIP systems

This paper studies energy-aware capacity configuration in two-class CONWIP systems with changing product mix. The goal is to select the number of active parallel machines at each station to minimise power consumption while meeting a throughput target. Because each capacity vector alters class throughputs, visit intensities, and local arrival mixes, the analytical aggregation must be rebuilt for every candidate configuration. The proposed configuration-specific evaluator adapts EMDA-style aggregation by updating station arrival mixes through a multiclass fixed point, reconstructing an equivalent single-class network, and solving a single-class MDA recursion at the total CONWIP population. Embedded in a hybrid search procedure, it enables rapid capacity comparisons without simulation during optimisation. Tests on structured scenarios, random instances, and a manufacturing-inspired case yield throughput errors of about 1%, strong ranking preservation, and reliable feasibility classification in the tested two-class region. The reconstruction is also decision-relevant: at the operational throughput tolerance, it changes the selected capacity vector in about one in five realistic instances within the exhaustively enumerated (Formula presented.) divergence pool. Among 39 DES-adjudicated divergent instances, no frozen rule found a better configuration, while in 24 cases it selected one that DES classified as infeasible.