Existing redundancy allocation and maintenance optimization models commonly neglect imperfect monitoring and maintenance conditions and often rely on sequential decision approaches. This study proposes a joint optimization framework for redundancy allocation and CBM in heterogeneous repairable multi-state weighted k-out-of-n systems. The objective is to minimize life-cycle cost while satisfying availability requirements under imperfect monitoring and maintenance conditions. Semi-Markov processes and universal generating functions are jointly used to model degradation, false alarms, missed detections, maintenance delays, and imperfect preventive actions. An adapted GA identifies high-quality feasible solutions under strict availability requirements. Results show that explicitly modeling imperfections reduces life-cycle cost by about 13%, while a sequential scheme increases cost by about 7% compared with the joint formulation. Sensitivity analyses reveal that demand and repair duration strongly affect performance and require joint adaptation of thresholds and redundancy.