STATISTICAL-INFERENCE OF UNSYMMETRICAL SILO PRESSURES FROM COMPREHENSIVE WALL STRAIN-MEASUREMENTS

A thin cylindrical shell structure which is subjected to local or unsy mmetrical loading often displays a very complex pattern of response, i nvolving multiple alternative potential failure conditions in differen t parts of the structure. The loading may therefore need to be defined with great precision. In the field of silo structures, it is widely r ecognised that such local loads often exist, but experimental observat ions of the patterns of load are very difficult to obtain because of t he expense of instrumentation and the need to use full-scale testing t o avoid granular solid scale errors. This paper presents a newly devel oped technique which permits these local unsymmetrical load patterns t o be determined in a much more cost-effective way. In addition, becaus e the loading is deduced from the structural response, the method has an inherent robustness in that when the deduced loadings are generalis ed and used to predict a structural response, it is more likely to be close to the real response. The same cannot be said for loading patter ns deduced from single discrete observations of loading with imaginati ve interpolations between them, which form the basis of most current d esign rules. The paper describes a rigorous procedure for inferring th e complete pressure distribution from a large body of strain observati ons on the silo wall. The method is outlined and a simple practical ex ample, involving unsymmetrical loads, is used to explore the effect of observation errors on the inferred pressures. A sample set of pressur es in a specially built full-scale test silo under eccentric solids di scharge is also derived. (C) 1998 Elsevier Science Ltd. All rights res erved.