Research On Maintenance Invest Strategies In Taffic Signal System

Light Emitting Diode (LED) Traffic Signals have become an efficient and effective alternative to traditional incandescent signals. The three main advantages of LED signals are low power consumption (10-22 W), long life, as long as 7 to 10 years and brighter. In addition to the low energy usage, the long life of LED signal means low maintenance costs, which makes LED signals a worthwhile investment and also environment-friendly.Despite their excellent performance, several barriers hinder more rapid retrofit in Hiroshima such as high retrofit cost and capital constraints. Thus, one of the most significant challenges in retrofit strategy is the decision-selecting which intersection to signalized with LEDs, which to keep on the table. We provide a method for selecting intersections not through traditional effectiveness measures like cost and performance, but instead through a quantitative analysis of the embedded uncertainty in each potential intersection.In this thesis, we evaluate the cost savings in LEDs retrofit program in Hiroshima, Japan, based on Life-Cycle Cost Analysis. Then, we propose a methodology for allocating capital across intersections using mean-variance analysis integrating structure learning method. We take 10 intersections in Hiroshima with their 8 years accident numbers into account, in which portfolios of retrofit can be analyzed and optimized. We propose an effective retrofit strategy where risk, measured by the variance in accident numbers, is considered together with accident mean. Our analysis of trade-off between accident numbers mean and variance employs mean-variance analysis and meta-controlled Boltzmann machine which is considerably efficient as the number of intersections dramatically increases.Besides, parameter K which included in Meta-controlled Boltzmann machine is evaluated and provides an evidence to select optimal strategies.