An intelligent procedure for updating deformation prediction of braced excavation in clay using gated recurrent unit neural networks

This paper aims to establish an intelligent procedure that combines the observational method with the existing deep learning technique for updating deformation of braced excavation in clay. The gated recurrent unit (GRU) neural network is adopted to formulate the forecast model and learn the potential rules in the field observations using the Nesterov-accelerated Adam (Nadam) algorithm. In the proposed procedure, the GRU-based forecast model is first trained based on the field data of previous and current stages. Then, the field data of the current stage are used as input to predict the deformation response of the next stage via the previously trained GRU-based forecast model. This updating process will loop up till the end of the excavation. This procedure has the advantage of directly predicting the deformation response of unexcavated stages based on the monitoring data. The proposed intelligent procedure is verified on two well-documented cases in terms of accuracy and reliability. The results indicate that both wall deflection and ground settlement are accurately predicted as the excavation proceeds. Furthermore, the advantages of the proposed intelligent procedure compared with the Bayesian/optimization updating are illustrated.     

TRD水泥土搅拌墙在武汉地区深基坑工程中的应用

武汉地区深基坑落底式止水帷幕目前主要采用地下连续墙直接施工至基岩的方式。受地下连续墙施工工艺及垂直度控制等因素的影响,该方式往往不同程度的存在渗漏水现象,且造价相对较高。本文通过武汉长江航运中心大厦深基坑工程设计实例,初步研究了TRD水泥土搅拌墙作为落底式止水帷幕在武汉地区一级阶地土层中的施工可行性、成墙质量及渗透性情况。经过技术方案对比,采用TRD水泥土墙落底式止水帷幕,止水效果好,可先行施工,缩短了基坑支护结构施工工期,且较地下连续墙落底经济,为武汉地区落底式止水帷幕的设计提供一种新思路。     

Identifying the cutting tool type used in excavations using neural networks

The paper presents results of preliminary research on utilising neural networks to identify excavating cutting tool’s type used in multi-tool excavating heads of mechanical coal miners. Such research is necessary to identify rock excavating process with a given head, and construct adaptation systems for control of excavating process with such a head.     

Modeling construction processes using artificial neural networks

The paper evaluates a neural network approach to modeling the dynamics of construction processes that exhibit both discrete and stochastic behavior, providing an alternative to the more conventional method of discrete-event simulation. The incentive for developing the technique is its potential for (i) facilitating model development in situations where there is limited theory describing the dependence between component processes; and (ii) rapid execution of a simulation through parallel processing. The alternative ways in which neural networks can be used to model construction processes are reviewed and their relative merits are identified. The most promising approach, a recursive method of dynamic modeling, is examined in a series of experiments. These involve the application of the technique to two classes of earthmoving system, the first comprising a push-dozer and a fleet of scrapers, and the second a loader and fleet of haul trucks. The viability of the neural network approach is demonstrated in terms of its ability to model the discrete and stochastic behavior of these classes of construction processes. The paper concludes with an indication of some areas for further development of the technique.     

敏感环境深基坑工程TRD工法等厚度水泥土搅拌墙设计与实践

淮安雨润国际广场工程为地处市中心的超大面积深基坑工程,该工程周边环境保护要求极高,基坑开挖深度范围内及基底以下巨厚的粉土、粉砂层富含地下水,对基坑工程隔水设计提出了极高的要求。为确保基坑隔水效果、保护周边环境,本工程在江苏地区首次采用了由TRD工法构建的等厚度水泥土搅拌墙隔水帷幕,有效的隔断了地下水,确保了基坑和周边环境安全。详细介绍了淮安雨润国际广场基坑工程的设计,并着重阐述了TRD工法等厚度水泥土搅拌墙隔水帷幕在本工程中的实施效果。     

软土地区深基坑变形控制设计实践与分析

以上海市瑞金医院普通病房综合楼基坑工程为案例,介绍了在环境保护要求较高的条件下采用地下连续墙加多道内支撑等设计、施工措施来控制基坑变形,通过对施工过程中周边管线及建筑、墙体变形等监测数据的分析显示,这些变形控制措施达到了预期的效果,可为类似的深基坑围护设计和信息化施工提供参考。     

软土深基坑预应力重力式围护结构的设计与分析

本文讨论了软土基坑工程中无支撑围护结构的主要型式及其特点,并对新型围护工法“垂直预应力锚杆补偿重力式围护结构”的主要工程特性进行了分析。通过工程实测和二维数值分析,证明该工法应用于软土深基坑的可靠性和优越性。     

Prediction of rotation capacity of wide flange beams using neural networks

This study proposes Neural Networks (NN) as a new approach for the estimation and explicit formulation of available rotation capacity of wide flange beams. Rotation capacity is an important phenomenon which determines the plastic behaviour of steel structures. Thus the database for the NN training is directly based on extensive experimental results from literature. The results of the NN approach are compared with numerical results obtained by a specialized computer. Available rotation capacity is also introduced in a closed form solution based on the proposed NN model. The proposed NN method is seen to be more accurate than numerical results, practical and fast compared to FE models.     

Prediction of geological hazardous zones in front of a tunnel face using TSP-203 and artificial neural networks

This research aims at improving the methods of prediction of hazardous geotechnical structures in the front of a tunnel face. We propose and showcase our methodology using a case study on a water supply system in Cheshmeh Roozieh, Iran. Geotechnical investigations had previously reported three measurements of the newly established method of TSP-203 (Tunnel Seismic Prediction) along 684 m of the 3200 m long tunnel up to a depth of 600 m. We use the results of TSP-203 in a trained artificial neural network (ANN) to estimate the unknown nonlinear relationships between TSP-203 results and those obtained by the methods of Rock Mass Rating classification (RMR – treated here as real values). Our results show that an appropriately trained neural network can reliably predict the weak geological zones in front of a tunnel face accurately.     

Predicting settlement along railway due to excavation using empirical method and neural networks

More and more excavation projects are being performed near existing buildings and structures due to large-scale urban construction, in which the excavation unavoidably causes settlement and potential danger to the surrounding construction and buildings. For linear traffic facilities parallel to the excavation, the settlement profile parallel to the excavation, namely, the settlement along the traffic line, should also be considered. Moreover, the precise control of the differential settlement along the traffic lines also plays a very important role. Thus, it is necessary to establish a quick prediction model, which is able to consider both vertical and parallel settlement profiles, using the basic information on the excavation. Based on the large amount of field data, the characteristics of the settlement profiles are analyzed. A simplified empirical method is proposed; it is established based on the Rayleigh and Gauss distribution functions for predicting the ground settlement along railways induced by an excavation. Meanwhile, back-propagation neural networks are also used to predict the settlement behavior. A comparison between the predicted results and the monitoring data is given to verify the feasibility of the proposed method. A good agreement indicates that the proposed method can be employed to predict the settlement along railways due to an adjacent excavation.     

Prediction of web crippling strength of cold-formed steel sheetings using neural networks

This study considers the use of neural networks (NNs) to predict the web crippling strength of cold-formed steel decks. Web crippling is critical for slender webs as in the case of trapezoidal sheetings which are widely used in roofing applications. The elastoplastic behaviour of web crippling is quite complex and difficult to handle. There is no well established analytical solution due to complex plastic behaviour. This leads to significant errors in various design codes. The objective of this study is to provide a fast and accurate method of predicting the web crippling strength of cold-formed steel sheetings and to introduce this in a closed-form solution which has not been obtained so far. The training and testing patterns of the proposed NN are based on well established experimental results from literature. The trained NN results are compared with the experimental results and current design codes (NAS 2001) and found to be considerably more accurate. Moreover, a trained neural network gives the results significantly more quickly than the design codes and finite element (FE) models. The web crippling strength is also introduced in closed-form solution based on the parameters of the trained NN. Extensive parametric studies are also performed and presented graphically to examine the effect of geometric and mechanical properties on web crippling strength.     

洪凝河特大桥深基坑挖孔桩板墙结构防护技术

结合洪凝河特大桥基坑现场实际施工情况，在综合考虑安全、经济等各方面因素后，介绍了施工钢筋混凝土挖孔桩板墙结构进行防护的方案，取得了较好的施工效果，对类似工程具有参考价值。     

CSM工法在软土地区深基坑承压水控制中的应用

以上海前滩33-01地块项目深基坑工程为背景,介绍了软土地区深厚承压水含水层的水文地质条件下,在上海地区首次采用50m深CSM工法等厚度水泥土搅拌墙作为落底式止水帷幕,隔断承压水含水层,保护基坑及周边环境的安全。通过CSM工法钻芯取样结果以及基坑监测数据的分析表明,基坑工程采取CSM工法等厚度水泥土搅拌墙,隔水效果良好,有效的控制了基坑降承压水对周边保护对象的影响,可以为类似工程提供一定的参考。     

Prediction of TBM jamming risk in squeezing grounds using Bayesian and artificial neural networks

This study presents an application of artificial neural network(ANN) and Bayesian network(BN) for evaluation of jamming risk of the shielded tunnel boring machines(TBMs) in adverse ground conditions such as squeezing grounds.The analysis is based on database of tunneling cases by numerical modeling to evaluate the ground convergence and possibility of machine entrapment.The results of initial numerical analysis were verified in comparison with some case studies.A dataset was established by performing additional numerical modeling of various scenarios based on variation of the most critical parameters affecting shield jamming.This includes compressive strength and deformation modulus of rock mass,tunnel radius,shield length,shield thickness,in situ stresses,depth of over-excavation,and skin friction between shield and rock.Using the dataset,an ANN was trained to predict the contact pressures from a series of ground properties and machine parameters.Furthermore,the continuous and discretized BNs were used to analyze the risk of shield jamming.The results of these two different BN methods are compared to the field observations and summarized in this paper.The developed risk models can estimate the required thrust force in both cases.The BN models can also be used in the cases with incomplete geological and geomechanical properties.     

Benefit of splines and neural networks in simulation based structural reliability analysis

Simulation based algorithms are often used to calculate an accurate value for the system reliability of complex systems. These concepts are very appealing because of their inherent simplicity. This is even more emphasized when used in combination with an implicit limit state function, for which the outcome is calculated by means of a finite element analysis. One of the major disadvantages however is the large number of simulations required to obtain an accurate estimate of the failure probability. This might result in an unrealistic processing time, making the method unusable for practical purposes. To meet this disadvantage, reliability analysis based on simulation methods in combination with an adaptive low order polynomial response surface are often used. The applicability has been demonstrated extensively. Ideally, no functional form is preset. The objective of this paper is to further increase the efficiency of simulation based reliability algorithms. Therefore the low order polynomial response surfaces are extended using neural networks and splines. The reliability framework is presented, compared with traditional response surface methods and commented extensively. The overall behaviour of the technique is addressed referring to several benchmark examples.     

Structural optimization for performance-based design in earthquake engineering: Applications of neural networks

This work addresses an approach to performance-based design in the context of earthquake engineering. The objective is the optimization of the total structural cost, under constraints related to minimum target reliabilities specified for the different limit states or performance requirements. The problem involves (1) the use of a nonlinear, time-stepping dynamic analysis to investigate the responses of relevance to the performances’ evaluation and (2) the integration of the responses into measures of damage accumulated during the earthquake. The random responses are deterministically obtained for different combinations of the design parameters and the intervening random variables, of which some are associated with the structure and some with the earthquake characteristics. The approach uses a neural network representation of the responses and, for each one, the variability associated with different earthquake records is accommodated by developing two networks: one for the mean response over the records, and another for the corresponding standard deviation. The neural network representation facilitates the estimation of reliability by Monte Carlo simulation, and the reliability achieved in each performance level, for a specific combination of the design parameters, is itself represented with a neural network. This is then used within an optimization algorithm for minimum total cost under reliability constraints. An application example uses a reinforced concrete, multi-storey plane structure with seismic demands corresponding to the city of Mendoza, Argentina.     

Backfilling missing microbial concentrations in a riverine database using artificial neural networks

Predicting peak pathogen loadings can provide a basis for watershed and water treatment plant management decisions that can minimize microbial risk to the public from contact or ingestion. Artificial neural network models (ANN) have been successfully applied to the complex problem of predicting peak pathogen loadings in surface waters. However, these data-driven models require substantial, multiparameter databases upon which to train, and missing input values for pathogen indicators must often be estimated. In this study, ANN models were evaluated for backfilling values for individual observations of indicator bacterial concentrations in a river from 44 other related physical, chemical, and bacteriological data contained in a multi-year database. The ANN modeling approach provided slightly superior predictions of actual microbial concentrations when compared to conventional imputation and multiple linear regression models. The ANN model provided excellent classification of 300 randomly selected, individual data observations into two defined ranges for fecal coliform concentrations with 97% overall accuracy. The application of the relative strength effect (RSE) concept for selection of input variables for ANN modeling and an approach for identifying anomalous data observations utilizing cross validation with ANN model are also presented.     

Modelling optimal risk allocation in PPP projects using artificial neural networks

This paper aims to establish, train, validate, and test artificial neural network (ANN) models for modelling risk allocation decision-making process in public-private partnership (PPP) projects, mainly drawing upon transaction cost economics. An industry-wide questionnaire survey was conducted to examine the risk allocation practice in PPP projects and collect the data for training the ANN models. The training and evaluation results, when compared with those of using traditional MLR modelling technique, show that the ANN models are satisfactory for modelling risk allocation decision-making process. The empirical evidence further verifies that it is appropriate to utilize transaction cost economics to interpret risk allocation decision-making process. It is recommended that, in addition to partners' risk management mechanism maturity level, decision-makers, both from public and private sectors, should also seriously consider influential factors including partner's risk management routines, partners' cooperation history, partners' risk management commitment, and risk management environmental uncertainty. All these factors influence the formation of optimal risk allocation strategies, either by their individual or interacting effects.     

Predicting uniaxial compressive strength of serpentinites through physical,dynamic and mechanical properties using neural networks

The uniaxial compressive strength(UCS)of intact rock is one of the most important parameters required and determined for rock mechanics studies in engineering projects.The limitations and difficulty of conducting tests on rocks,specifically on thinly bedded,highly fractured,highly porous and weak rocks,as well as the fact that these tests are destructive,expensive and time-consuming,lead to development of soft computing-based techniques.Application of artificial neural networks(ANNs)for predicting UCS has become an attractive alternative for geotechnical engineering scientists.In this study,an ANN was designed with the aim of indirectly predicting UCS through the serpentinization percentage,and physical,dynamic and mechanical characteristics of serpentinites.For this purpose,data obtained in earlier experimental work from central Greece were used.The ANN-based results were compared with the experimental ones and those obtained from previous analysis.The proposed ANN-based formula was found to be very efficient in predicting UCS values and the samples could be classified with simple physical,dynamic and mechanical tests,thus the expensive,difficult,time-consuming and destructive mechanical tests could be avoided.