<p>This study develops a robust and resilient multi-objective decision-support framework for the integrated order acceptance, production planning, scheduling, and preventive maintenance (OAPPS-PM) problem under uncertainty. While most existing models address these decisions separately or only partially integrate them, this research proposes a unified Multi-Objective Mixed-Integer Nonlinear Programming (MOMINLP) formulation that concurrently minimizes total cost and delivery delay while maximizing order fulfillment. The model captures real-world manufacturing complexities, including machine availability, material lead times, task precedence, and preventive maintenance scheduling. Unlike conventional robust optimization approaches that rely on fixed uncertainty sets, a scenario-based adaptive uncertainty modeling scheme is introduced, allowing dynamic representation of stochastic variations in processing times, lead times, and equipment reliability. Risk aversion and infeasibility penalties are incorporated to flexibly control trade-offs between robustness and performance. To address computational challenges, two complementary solution strategies are developed, the Enhanced Augmented ε-Constraint (EAE-ε) method to generate a well-balanced Pareto frontier and a Hybrid Adaptive Large Neighborhood Search with Scenario Grading and Surrogate Evaluation (HALNS-SGSE) that utilizes XGBoost surrogate models for rapid scenario evaluation for large-scale instances. Application to a real-world battery manufacturing case demonstrates that HALNS-SGSE achieves cost reductions of up to 7.3%, delay reductions of 11.6%, and order fulfillment improvements of 6.2% compared to baseline heuristics, while maintaining average infeasibility below 0.05 across 30 stochastic scenarios. The proposed framework thus provides an effective and generalizable tool for resilient decision-making in uncertain production environments.</p>