Research On Key-parameter Experiments And PLS-BP Correction Approach: LRAD Measurements Inside Pipes

With the implementation of nuclear facility decommissioning and waste treatment project, ever-increasing waste pipes were internally contaminated with various and often undetermined radio nuclides, which need to be measured validly. As alpha particles have a very short range and weak penetration in matter, it's hard to externally detect alpha contamination on the inside surface of pipes. Currently acceptable alpha detection technologies (Wipe measurement or probe ones) present some limitations. For alpha-particle detection requirements of pipe interiors and irregular surfaces, how to effectively monitor radioactivity inside these pipes has become difficult point.LRAD (Long Range Alpha Detector), a highly sensitive method that measuring alpha particle radioactivity is proposed, which can achieve the object surface contamination of non-destructive, fast and long-range detection, can measure in any space where air can penetrate. Thus has a broad prospect of application.From 1991 to now, the application of LRAD technology mostly includes waste pipe surface contamination detection, real time monitoring inner surface, soil surface pollution and radon. Up to the 20th, a few domestic institutions began to carry out LRAD research as well, characterizing the pipe-end monitoring system based on LRAD technology, which gains certain experience. However, comparing with foreign counterparts, basic data on measuring alpha contamination inside pipes seriously scares, further basic research work and technology development should be strengthened, such as conducting key parameter impact characteristic research on measuring system, including parameters of source-to-detector distance, test pipe diameter, pipe length, air flow and pipe geometry, etc. In this way the LRAD measurement system can be effectively and comprehensively promoted in fields of decommissioning of nuclear facilities.In radioactive measurement field, nonlinear problem is the key to affect system accuracy. LRAD measuring pipe interiors, susceptible to multi-parameter, also results in serious nonlinear problem between the ions collected and alpha source strength. The factors causing the nonlinear phenomena make large deviation only based on the linear calibration, like distribution for goal alpha pollutants, test pipe diameter, test pipe length, pipeline geometric, air flow, etc. Therefore, for effectively measuring pipe interiors, exploring the nonlinear impact factors on LRAD measuring system, conducting nonlinear correction approach research will be helpful to improve the LRAD analysis precision, also be an important research direction.According to the key questions above to be solved,on the basis of LRAD investigation situation at home and abroad, the emphasis is placed on the key parameter experiment of LRAD monitoring pipe interiors and nonlinear correction method research of partial least square BP network, which is to independently developed LRAD measuring system as a tool, to ²³⁹Pu alpha sources of different surface activity as measuring objects.The major research contents and results are as follows:(1) Systematically conduct the key-parameter impact feature experiment for LRAD monitoring pipe interiors.Basing on the research achievements of LRAD measuring technology at home and abroad, by simulating pipeline, placing alpha source inside the pipe system, the response features were characterized among source orientation inside pipe, source-to-detector distance, test pipe diameter, pipe length, air flow and pipe geometry, and the impact factors were found out.(2) Establish parameter impact characteristic equationsAccording to analyze impact features of single-parameter and double-parameter change on measuring system, the thesis adopts reasonable mathematical fitting methods, respectively establish impact characteristic equations on the response versus radioactive sources, distance, test pipe length, test pipe diameter and air flow. Research shows that source activity and distance presenting linear characteristics on system detecting efficiency, polynomial characteristics for test pipe diameter and length, exponential change for air speed and air flow, no influence for pipe geometry and the source placement inside pipes, basically.(3) Use the partial least square BP network to comprehensive nonlinearly predict experimental data obtained.Based on previous research of key-parameter impact characteristics, this thesis considered each impact parameter as an impact factor, proposed partial least-square BP neural network analysis method to establish a dynamic nonlinear prediction model of pipe interior LRAD measuring system, namely the PLS-BP network model. The model first preprocess experimental data through the partial least-square thoughts that solved the defects of BP network, and then through BP network model mapping the nonlinear relation and between key parameters and measured results in measuring system, which is of functional approximation ability with arbitrary precision, can be fixed automatically. The results show that the model is simple and effective by combining partial least-squares regression thought with the BP neural network forecasting model. The relative error predicted can control at 6%, which greatly reduces nonlinear effect.The thesis has developed a fundamental research work, and also made some innovation on methods and techniques, whose final results achieved the purpose of effective nonlinear calibration, and establish a set of complete nonlinear correction method.Generally, the innovations of this paper are the following three points:(1) Firstly conduct simulated experiment at home, systematically carrying out basic research with the independently developed LRAD monitoring system.(2) Obtain multi-parameter impact characteristic to the results, and establish mathematical characteristic equations.(3) Propose PLS-BP network analyzing method, and establish nonlinear prediction model between the impact parameters and measurement results.The thesis using the LRAD technology, through the impact characteristics of key-parameter experiments, obtain mass valuable data, systematically study multiple-parameter impact characteristics on measurement results including the system response versus test pipe diameter, length, air flow, detection distance and other factors. Finally, based on the prophase experimental results, use the PLS-BP network analyzing method to establish the correction model between the impact parameters and measurement result. Through the theory and application research of LRAD monitoring pipe interiors, more valuable data on alpha radioactive monitoring is accumulated, which contributes a lot on our country's nuclear facility decommissioning and environmental monitoring research.