Search

Every day, millions of people depend on the safe and reliable transportation of oil and gas through pipelines. Pipelines are historically a very safe means of transporting large quantities of oil, natural gas, fuels, and other hazardous materials. However, accidents of pipeline leakage have occurred frequently because of pipelines age, rot and third-party damage and destruction. Waste and pollution occur and result in such serious disasters as fire and explosions. A dependable leak detection system is important to promptly identify the occurrence of a leak in order to shut down the line, isolate the leak, initiate response actions, reduce the volume of the spill, and mitigate safety, environmental, and economic consequences of the release. Leak detection methods cover a wide spectrum of technologies and are based on a number of different detection principles. They vary from intermittent aerial inspections to hydrocarbon sensors to Supervisory Control and Data Acquisition (SCADA) based, real-time monitoring. Each approach has its strengths and weaknesses. This paper describes the capabilities and limitations of current leak detection systems. The operational principle, data and equipment requirements, strengths, weaknesses, and realistic performance limits for the leak detection methods are addressed in this paper.

Through the investigation on car-to-bicycle accidents of Beijing city, the characteristic impact parameters of human, bicycle and vehicle are collected. The impact speeds are classified by statistic results and law regulation. Above that, based on the Artificial Neural Network (ANN) method, relevant typical parameters are selected to build an estimate model on vehicle impact speed of car-to-bicycle accidents. The input layer of the model contains 45 nodes, including human parameters, bicycle parameters, car parameters, road parameters and other information. The output layer of the model represents the forecasting data and different classification results of vehicle impact speed respectively. Using the credible data from real accidents, the model can be trained and applied in vehicle-speed analysis and accident reconstruction.