Safety Monitoring Theories And Technologies For Earth Rockfill Dam

Earth rockfill dam (ERD) has the characters of remarkable parameter variability, complex assessment rules and circumstance uncertainty to the disease diagnosis, so its safety monitoring is more complicated than concrete dam and there are unique monitoring theories and methods for ERD. Theories and technologies on ERD studied in this paper are as following:(1) The error of inverse plumb system in ERD is analyzed. The anchor point stability is studied by FEM analysis with water level variation combined. The optimization of depth of inverse plumb in high ERD is settled. The modifying model of inverse plumb group is researched and the stability in contradiction to sensitivity is solved. As a result, stable benchmarks can be acquired without greatly deep reserve plumb, and the problem of over-range is settled.(2) An automatic three-dimensional monitoring system is constructed by using the advanced technology of total station instrument. The gray models for calculation of refraction coefficient are established, in which valid altitude modification is conducted, so the question of atmosphere refraction in three-dimensional monitoring is resolved.(3) The theories for unascertained data are brought in error process. The unascertained filtering (UF) is brought forth and the related parameters are investigated. Consequently, the method of UF is improved by examination and experiment therefore the improved unascertained filtering (IUF) is invented. Using this method, the problems of error hiding, normal data discarding and hard to differentiate error from abnormity existing in the traditional methods for detecting error are settled. IUF has the character of dynamic and real-time, and it has predominance in solving long series with wide span at numerical value from ERD. Application domain of unascertained data theory is extended, and a new approach to error process is acquired through the research.(4) There is shortage of hiding data in traditional methods for abnormality detecting. Based on swatch dividing theory and with feature of monitoring data from ERD considered, the result of abnormality detecting is described in four degrees by using three different test levels. Traditional detecting work depending too much on the data series selected, the concept of reliability is brought forward. The shortage is overcome by several data schemes consulted and correspondingreliability put to each scheme. The result of abnormality detecting is more all-sided and reasonable, and the conditions to correct and successful comprehensive assessment can be ensured.(5) The blind data theory is excerpted into the comprehensive assessment field. Based on the blind information concept, there is a conclusion that the response from safety monitoring in ERD is a kind of blind information. Then a method based on blind data theory for comprehensive assessment is put forward. The information collected by monitoring instruments and the patrolling information is expressed in blind data form, consequently the complex operation rules are predigested by deducing applied formula. With this method used, the shortage of absolute assessment results is overcome. The application of back promulgating neural network (BPNN) to comprehensive assessment in ERD is realized, and the NN is optimized based on the research of input/output models transforming and the operation using remaining error feedback. Compared the assessment solution of BP NN with blind data theory, some significant conclusions are acquired.