Modelling Working Performances Of Lateraly Loaded Masonry Panels Using CA/NN Techniques And Quantum Entanglement Principle

Masonry is one of the most widely used composite materials in engineering.But the existing analytic methods are difficult to achieve the accurate prediction of a masonry structure because of its complex working behaviour with great variability.Therefore,it has been not only necessary but also classic issue to improve the analytical methods or develop the innovative techniques for accurately predicting the working performances of masonry structures.This thesis presents a series of innovative methods for predicting the failure patterns and failure loads of masonry wall panels and wallets applying cellular automata(CA),neural networks(NNs)and quantum entanglement(QE)principle.The cellular automata(CA)model was built to predict the failure patterns of laterally loaded masonry wall panels with different configurations.Two methods were applied to calculate the state values to such as series and exponential state values form the state modes of wall panels.And two criteria were introduced to match the state similarity among the cells(zones)between/in the state modes.The comparative analysis of the matching criterions presented the importance of data occurrence order for computer systems and the problem of symmetry effect on the accuracy of prediction.Hence,a position corrector was introduced to minimise the symmetry influence.And a new method was developed to calculate the state values based on boundaries force difference.For the prediction of the failure loads of masonry wall panels,the NN mode was built in consideration of the panel's configuration information.The opening ratio was proposed as the input of the NN model,leading to the accurate prediction of the load bearing capacity of wall panels with openingsThe deep-learning NN and CA techniques was combined to predict the crack patterns of masonry panels with different configurations.Two NN models were tested using the panel's experimental data to find the hidden connections.One calculated the state values based on Moor CA model and the key factor was suggested to characterize the derivation of cracked zones.The other NN model was based on the von Neumann CA model which propagated the four boundary conditions of the wall panels through transition functions into all the zones(cells)on masonry pattern.Then the built NN models were trained by an amount of experimental data and individual parameters.Both NN modes improved the accuracy of prediction.Prediction of failure load for both solid and panels with an opening was established through the combination of NN and CA.Two methods were developed for this purpose.First,the NN used panels' length,height,thickness and opening ratio and Moore CA to build preliminary patterns of loads by predicting a load value for every cell on the CA model.From those patterns,using the criterions for evaluating cracking zone and failure load,the predicted failure load was derived.Second,a new CA model called openingcentric model was established.In this model the panels were divided into nine cells according to the position of an opening.After receiving the information of the nine cells,wallette strength and panel's configuration coefficient the NN predicted the failure load.Both methods built accurate predictions of failure load,also the training of the NN was much faster for the second method.The principle of quantum entanglement was tried to establish the techniques for predicting the failure patterns and failure loads of laterally loaded masonry wall panels.The Quasi-quantised(QQ)techniques also involved in cracking pattern's modelling,transmission effect and matching criterion.The QQ model implied that the entanglement degree could limit the choice of a base panel for accurately predicting crack patterns.The QQ formula was proposed to predict the failure loads of wall panels.To bypass the restrictions of entanglement degree,the relationship between the heights,lengths and thicknesses of base panel and predicted panels was analytically and mathematically derived.The research results achieved in this thesis contributed to the structural analysis theories.And the of accurate prediction capacity of the proposed CA,NN and QQ models could provide the reference to the improvement of design methods and lead to the anticipated engineering benefits.