Statistical Inference Methods For Characteristic Value Of Floor Live Load On Minor Sample

In the evaluation of structural reliability, the inference of building’s floor live loadis a very important work. It is important basis in reconstruction and reinforcementdesign of existing structures and important reference in determining the characteristicvalue of floor live load in national load code of proposed structures. During inferringthe characteristic value of floor live load, if the inferred numerical value is too large, itwill cause the waste of materials and increase the cost of construction, if the numericalvalue inferred is too small, it will cause the decrease of structural reliability and the lackof security. Hence it is very important to infer the characteristic value of building’s floorlive load reasonably an accurately. At present, the theory and method of the inference ofcharacteristic value of building’s floor live load in the condition of large sample ismature, national standard has given the recommended method of the inference ofcharacteristic value of building’s floor live load in our country. The recommendedmethod is based on the condition of large sample, but in the engineering practice, not allsample are large sample. In fact, because of various reasons, we can’t get sufficientsample size and this will lead to produce the minor sample. Then we won’t be able toinfer the characteristic value of building’s floor live load according to the method oflarge sample given, otherwise the result will be too inaccurate to meet the requirementsof the actual situation. And now there is no mature theory and method to infer thecharacteristic value of building’s floor live load in the condition of minor sample.For the unsettled small sample problem, this paper proposes two methods. The firstone is the estimation method that uses the fractile of the maximum value type I distribution to establish the inference method of the characteristic value of building’sfloor live load in the condition of the minor sample. It can be named liner regressioninference method. The second method is the bayesian inference method of inference ofcharacteristic value of building’s floor live load in the condition of minor sample, whichis based on the correlation theory of belief. This paper infers the formulas in thecondition of the variance is known and unknown, and gives the inferred results. Finally,based on the survey statistics data of a mall’s floor live load, the paper infer itscharacteristic value with three different methods, and compare and analyse the resultinferred, obtain the conclusion. Through the comparison and analysis it can be foundthat the linear regression method is relatively accurate, and the error of the bayesianmethod is bigger, but it is relatively conservative. At last, the paper give somerecommended value of some parameters in the two inference methods.Achievement of this paper provides two more reasonable methods for the inferenceof characteristic value of building’s floor live load under the condition of minor sample，it is good for getting reasonable inferred numerical value of the existing structure’scharacteristic value of the floor live load, and it provides another methods fordetermining the characteristic value of building’s floor live load of structures in design,it could also provide the reference on theory and method for the revision of nationalload code.