A Bayesian network-based approach to constructing a task risk management and optimisation model for university renovation works

Abstract

College renovation projects face the triple management dilemma of mixed personnel, hidden building risks, and lack of dynamic data correlation, and the traditional methods have the defects of coarse risk identification granularity, lagging emergency response, and inaccurate multi-objective optimisation. Aiming at this, a dynamic risk optimisation model integrating multimodal biometrics and Bayesian network is proposed: based on iris/fingerprint bimodal authentication to achieve accurate tracking of personnel status (misidentification rate ≤ 0.001%), constructing 'personnel-equipment-environmental' dynamic topology, and adopting an improved Markov chain Monte Carlo algorithm to solve the problem of insufficient historical data (<200 sets). The modified Markov chain Monte Carlo algorithm is used to solve the parameter learning problem under the insufficient historical data (<200 groups), and combined with the non-dominated sorting genetic algorithm to generate the safety-cost-duration Pareto frontier solution set. In the renovation of a teaching building of a century-old university, the model improves the risk identification coverage from 63% to 91%, compresses the emergency response time from 4.7 hours to 1.2 hours, improves the resource utilisation rate by 37%, and saves 2.86 million RMB in cumulative cost. The innovativeness is reflected in the multimodal biological data synchronous analysis architecture, dynamic conditional probability table correction mechanism, and low-power deployment scheme for edge computing terminals (<5W/node), which provides quantifiable and traceable decision support tools for highly-mixed renovation projects, and is certified by the National Centre for Quality Supervision and Inspection of Construction Engineering, and is of value to the industry for promotion.