A Study On The Demarcation Of High Slope Stability Of Step One Hydro-junction In Jingping By Artificial Neural Network

Jingping step one hydropower station locates in lower zone between gaugestation and Shoupa groove of big river bend where is in upstream of Yalong riverwhere is in southwester of Sichuan province. The hydropower station locates inborder land between Yanyuan county and Muli Li national autonomous county.The hydropower station dam design height is 305m. the normal impoundingwater height is 1880m. The equipped capacitor is 3,600MW, that equivalents 1.5folds generating capacity of Three Gorges hydropower station. The reservoircapacity is 77 hundred million cubic meter and the hydropower station is themost reservoir capacity at Yalong river drainage area. Now, Jingping step onehydropower station entered the construction excavation phase. The plan is thatthe river will being stopper flow in 2006 and the first electric generating set willyield electricity in 2012 and final completion in 2014.The river valley is narrow and the valley slopes are steep at the bank ofjunction zone. The deforming and breakage phenomena were appeared in someslopes that formed tension fracture rock mass. The deformable body that haddistinct boundary was formed in some slopes. The other some slopes had beenformed different scale landslide and slide. So the stability of high steep valleyslope is very urgent at junction zone.The junction zone slope is the slope in riversides that length is about 1.5 km.For the purpose of highlighting focal point, it is necessary that stability zoningbased on geologic examination. This stability zoning is qualitative studying. Forthe purpose of getting quantitative or semi quantitative studying results, thethesis studied the stability zone of slope at junction zone based on artificialneural network. The results were the foundation of subsequent studying, so thatis significance.In addition, the thesis adopted limiting equilibrium and numericalsimulation to study the stability of left bank slope at downstream side of V#mountain at junction in Jingping step one hydropower station. And the resultswere contrasted with the artificial neural network computed results. The studyresults were the basis of every engineering slope design, so it was significance.At the same time it supplied typical example of high slope study."General system science" and "geology process mechanics analyzing andquantitative assessment" guided the thesis to construct BP of artificial neuralnetwork and probability of artificial neural network adapting to judge thestability of slope. At the base of geology analyses the thesis studied the stabilityzone of high slope at junction in Jingping step one hydropower station and theresults that were computed by limiting equilibrium and numerical simulationwere contrasted with the artificial neural network computed results. At the sametime the thesis sufficient and systemic studied the stability of differentengineering condition of left bank slope at downstream side of V# mountain andright bank Pusiluo groove engineering slope. These all can be used as thestability studying results duration of construction and operation period, at thesame time these results supplied typical example of building other hydropowerstation and searched the study of high steep slope. So it was significance to buildJingping step one hydropower station and study slope engineering.The artificial neural network has extensive application in many fields. Ithas much application in the field of engineering geology too. But the method isfar immature and exists defect not only in theory but also in practical application.So it is very significance to study the method and to create a more rational andbetter utility artificial neural network.The thesis introduced the application situation of artificial neural networkto rock mechanics and the basic algorithm of BP artificial neural network. Thethesis also generalized various improved algorithm and used homotopyalgorithm to accelerate the velocity of slow algorithm. The BP artificial neuralnetwork with importation layer, implication layer and output layer wasconstructed by MATLAB6.x software. The number of Every layer neuron couldbe randomly designed in the BP net. 12 training algorithms, 5 initializationmethods of weight value, 5 transitive functions and 4 error formats wereincluded in the BP net. So let users randomly select different calculation way toconstruct the BP net of different structure and algorithm. The thesis selected thepattern data of the paper that was published in engineering geology evaluationmethod for slope stability based on artificial neural network in journal of rockand soil mechanics by Yuanyou Xia to train the BP net to compare with differentalgorithm in characteristics, merits and faults. The thesis also studied theselection of the number of implication layer neuron. The results indicated thatwhen the other condition was identical, if the multiple of the number ofimplication and importation layer neuron was n, then the net error was minimumif n was 1.0~2.0.At the same time, the probability of artificial neural network wasconstructed in the thesis. The same training and testing pattern data as the data ofBP net were used in the probability net. After training, it was known that the netcould perfectly class the 29 pattern data when esp≤0.4 and the correct ratio of theforecasted outcome of 7 pattern data was highest, moreover the ratio higher thanthe correct ratio of BP net, when 0.02≤esp≤0.4.In general geology analysis, artificial neural network, limiting equilibriumand numerical simulation were combined to study the high slope stability and tozone as stability zone at Junction in Jingping Step One Hydropower Station inthe thesis. In the process, the BP net with 15-31-7 structure and the probabilitynet with 15-7 structure were constructed in the thesis and the high slope stabilityat Junction in Jingping Step One Hydropower Station was forecasted by the net,moreover the thesis compared the results of artificial neural network withordinary computing to think that the results of artificial neural network wasreliable. So the thesis obtained key conclusions, such as:1. The thesis analyzed and generalized the slope structure type andfundamental mode of slope deformation and failure. Moreover the thesis zonedthe slope at junction as stability zone by the geology analysis way according tothe factors of landform, layer rock character, slope structure, the phenomena ofslope deformation and failure and engineering environment.2. By means of numerical simulation to analyze seepage field of IV#-VI#mountain in the condition of rapid flood, the thesis obtained some preliminarycognitions, such as: ①The groundwater was enrich at toe in the engineeringslope. In the mountain hydraulic gradient was gentle and majority rock mass wasover the groundwater table to become the unsaturated zone. In the condition ofrapid flood, the groundwater table all rose at the different raininess zone but therising degree was different. In general the degree was deepest, the degree at rightbank was less to left bank and the groundwater table was less 1700m at the leftbank deep joints zone in 48 hours. ②In the different raininess zone thegroundwater table rising degree was different at the left bank deep joints zone.③At the left bank deep joints zone the groundwater table persistently rose in theprocess of rapid flood time and integrated rose 9.72~43.01m in 24 hours, theaverage value was 0.40~1.79m/h. It has the tendency that the groundwater tablerising degree became shallower in the process of time.3. The RSE value of BP net was studied in the thesis and the weightconception of different input factors to whole output outcome was introduced atfirst based on the RSE value. The weight can be not only compared with theothers (example fuzzy evaluation and so on) but also used to judge the rationalityof the pattern data.4. The thesis separately forecasted the slope stability at junction by BP netand probability net and the less value in the two type results was used as theultimate result. After the analyzing result it was known that the forecastedmountains stability of artificial neural network was the same as the slopestability zone in the third chapter, so the slope stability zone was rational.5. It was thought that the integral deformation destabilization model maybesliding and fracturing at VI#-IV# mountain in the condition of rapid floodthrough the engineering geology deep analysis and analogy study. That is to saythat the rock mass with resistance effect was softened at toe to deform with thecharacter of side extrusion and lead to deteriorate the rock mass quality andfurther lead to the rock mass with resistance effect being sheared and the rockmass in middle and upper sliding by f9 fault. This was a whole deforming andfailure model. It was belong to the problem of deforming failure.6. It was known that the range influence of stress distribution was finiteafter training cave excavation by numerical simulation. The deformation valuewas only millimeter grade on the slope at VI#-IV# mountain that indicated theinfluence of stability was lighter for the reason of excavation.7. After excavating left training cave exit slope, the tensile stressdistribution range was expanded adjacent to cutting line and the value was less0.5MPa. The influence of stress distribution wasn't significant after excavatingcave exit slope. The f2 fault appeared in the range of cave exit slope excavationline. After excavation, the rock mass between f2 fault and roof in adjacent tocave exited bigger deformation that indicated the stability condition of the rockmass was worse, so in the course of construction it should be monitored andsupported in time.8. In the course of left training cave exit slope excavation, by means ofanalysis the character of the slope tensile stress distribution above cutting line atA1 section. It indicated that the tensile stress zone was occurrence at the worstweathered zone above cutting line and the stress was disadvantage to the rockmass stability on the slope, so it maybe partially collapse.9. In the direction of left training cave exit axis, the phenomena of tensilestress concentration was occurrence in adjacent of slope and above the cave roofby the means of analysis the character graph of the tensile stress distribution. Themaximum value was about 1MPa and the range of influence was abovetraining cave exit roof about 50~60m. The stress was disadvantage to the rockmass stability above training cave exit axis. As excavating the critical planegradually formed that was disadvantage to the cave exit slope.10. In the condition of rapid flood there were two factors of beingdisadvantage to the whole stability of VI#-IV# mountain: Deep joints zone werefilled water to increase load and the rock mass with resistance effect was relativethin at toe, moreover the rock mass maybe deteriorated in the condition of rapidflood. The two factors maybe lead to part and superficial slope collapse. Themain deformation destabilization model at superficial slope was bending-slidingand sliding-fracturing.11. Because limiting equilibrium only considerate the sliding strengthfactor of failure plane and artificial neural network can considerate more factors,the latter results would be conservative and more reliable at the same time.Moreover the results of numerical simulation computer were same as artificialneural network that could also illustrate the reliability of artificial neuralnetwork.