Cost Control in EPC Public Works Using a System Dynamics Model Embedded with Intuitionistic Fuzzy Reasoning: A Case Study of the Urumqi Civic Center

Cost control in engineering, procurement, and construction (EPC) public works is shaped by interacting drivers, nonlinear feedback, and qualitative judgments. Existing studies usually apply the three relevant method families separately: DEMATEL-ISM maps causal structure but does not propagate hesitation-aware expert judgments into cost trajectories; fuzzy systems represent uncertainty but commonly lack a verified causal hierarchy; and system dynamics (SDs) capture dynamic accumulation but often rely on crisp inputs. The resulting absence of a traceable causal screening to uncertainty to dynamic cost link is the specific gap addressed in this study. We therefore develop a transparent three-stage pipeline combining the Decision-Making Trial and Evaluation Laboratory–Interpretive Structural Modeling (DEMATEL-ISM) method, triangular intuitionistic fuzzy reasoning (TIFR), and SDs. DEMATEL-ISM identifies the causal hierarchy; TIFR represents membership, non-membership, and hesitation in design complexity and human–technology synergy judgments; and SDs evaluate stage-specific cost trajectories. Recalculation from the supplied 17 × 17 direct influence matrix produced a six-level hierarchy in which senior management decision-making capability and the level of integration occupy the two deepest driving levels. For the Urumqi Civic Center case, the baseline terminal cost absolute percentage error was 0.492%. A coordinated intervention scenario shifted the simulated terminal cost by CNY 12.1243 million (6.1%) relative to the baseline; this is a model-based scenario difference, not an observed project saving. Integration had the largest simulated effects on design and transportation costs, whereas senior management decision-making capability had the largest effects on procurement and construction costs. Security cost curves showed a complementary pattern between managerial capability and workers’ professional competence, but no statistical interaction effect is claimed. The framework is intended for within-case scenario comparison and intervention prioritization; multi-project and time-series validation remains necessary.