Service Load Research Articles

Toward efficient resource utilization at edge nodes in federated learning

Federated learning (FL) enables edge nodes to collaboratively contribute to constructing a global model without sharing their data. This is accomplished by devices computing local, private model updates that are then aggregated by a server. However, computational resource constraints and network communication can become a severe bottleneck for larger model sizes typical for deep learning (DL) applications. Edge nodes tend to have limited hardware resources (RAM, CPU), and the network bandwidth and reliability at the edge is a concern for scaling federated fleet applications. In this paper, we propose and evaluate a FL strategy inspired by transfer learning in order to reduce resource utilization on devices, as well as the load on the server and network in each global training round. For each local model update, we randomly select layers to train, freezing the remaining part of the model. In doing so, we can reduce both server load and communication costs per round by excluding all untrained layer weights from being transferred to the server. The goal of this study is to empirically explore the potential trade-off between resource utilization on devices and global model convergence under the proposed strategy. We implement the approach using the FL framework FEDn. A number of experiments were carried out over different datasets (CIFAR-10, CASA, and IMDB), performing different tasks using different DL model architectures. Our results show that training the model partially can accelerate the training process, efficiently utilizes resources on-device, and reduce the data transmission by around 75% and 53% when we train 25%, and 50% of the model layers, respectively, without harming the resulting global model accuracy. Furthermore, our results demonstrate a negative correlation between the number of participating clients in the training process and the number of layers that need to be trained on each client’s side. As the number of clients increases, there is a decrease in the required number of layers. This observation highlights the potential of the approach, particularly in cross-device use cases.

Риск-ориентированный анализ безопасности и защищенности промышленных объектов на основе характеристик прочности и ресурса

The concept of computation and experimental determination of safety and protectability of industrial facilities on parameters of risk of initiation of refusals, accidents and disasters taking into account possibility of regular and emergency situations is considered. The scientifically substantiated, normalized and controlled parameters of risks estimated by criteria of strength, life-time and survivability are its basis. Risks at the same time are estimated on probabilities of initiation of emergency situations and losses accompanying them at their interdependence. It is shown that generally safety is defined by ability of installations not to pass into emergency and catastrophic states with drawing significant damages to the person, a technosphere and environment; the risk is defined by probability of initiation on installation of adverse situations and possible losses from these situations; protectability is defined by ability of installation to resist to initiation and an evolution of adverse situations (refusals, accidents, disasters) in regular and emergency conditions. The new perspective direction for maintenance of conditions of safe service of industrial facilities is that which initially forms the level of their protectability, and this protectability is defined by the formed and acceptable risks. In this case the solution of basic problems of protectability, safety and risk has to be guided surely by the solution of problems of strength, a life-time, survivability in their direct interaction This circumstance is based on a quantitative estimation of conditions of service damages accumulation and on the analysis of conditions of transition from regular to the damaged and limit states of installation taking into account criteria of strength, fatigue life and survivability. The solution of practical problems of maintenance of strength, life-time and safety within of strength, life-time and risk-focused design assumes determination of nominal and local stresses from service loadings in the most loaded zones with use of characteristics of mechanical properties of constructional materials. In the paper it is put and solved the new scientific and almost important problem about risk-focused definition, maintenance, increase and a standartization of complex safety and protectability of technosphere installations by criteria of the acceptable and controlled risks.

Capacity assessment of railway axle journal bearing addressing a localized rolling contact fatigue failure mechanism

Abstract A framework for assessing the capacity of railway axle journal bearings is presented after elaborating on the locale rolling contact fatigue (RCF) failure mechanism. Three phases are conducted, including the determination of bearing service loads, finite element analysis (FEA) on the localized RCF stresses of bearing components, and the life-related capacity assessment on the bearings and bearing components. Results prove that wheel-track (WT) contact forces inspected by an instrumented wheel set can be transferred into bearing service loads. Two-step FEA from the global wheel set to the local bearings is used to obtain localized RCF stresses at dangerous locations around the bearing contact regions. After determining the locations, a life-related capacity assessment is performed by evaluating equivalent RCF stresses and lives at the locations to determine the weakest bearing component. Typical practice is given in agreement with FEA. This method deals with the FEA of bearing structures, material cyclic stress-strain curves, and probabilistic RCF life curves. Therefore, this work greatly enhances the development of rail-bearing materials, innovative structures, and capacity assessment.

Optimization of Hanger Spacing of Steel Arch Bridges Using Dynamic Loads

Bridges are basic infrastructure that must be met to create regional connectivity in Indonesia. One type of bridge that is often used is a curved bridge, which has the advantages of high strength, attractiveness, aesthetics and economy. In order to accelerate the development of bridge infrastructure, an efficient innovation in curved bridge design is needed. The development of curved bridge structures to achieve efficient designs has received much attention in several decades. However, researchers have only focused on optimising the geometry variation of the arch height. Therefore, the aim of this research is to innovate the optimisation of the hanger spacing on the arch bridge structure. In order to obtain optimal results, a bridge model is carried out by varying the hanger spacing of the centre model with a hanger spacing ratio of (1.3 - 1.1 - 0.9 - 0.7), a flat model with a hanger spacing ratio of (1 - 1 - 1 - 1) and an edge model with a hanger spacing ratio of (0.7 - 0.9 - 1.1 - 1.3), so that from the three models, the effect of hanger location on three conditions is obtained. Each model is modelled in the SAP2000 software and given a bridge service load to obtain the internal forces and deflections that occur. The output of the internal force and deflection is then analysed to determine the effect of the location of the bridge service hanger. The serviceability of the bridge is also analysed by calculating the ratio between the weight of the bridge and the deflection that occurs. The results of the analysis show that the location of the hanger affects the performance of the arch bridge structure. The centre model bridge design produces the most efficient structural performance in resisting the compressive axial forces and moments that occur, and produces the least deflection. Meanwhile, the edge model will provide the most efficient structural performance in resisting tensile axial forces. By referring to the results of the bridge weight to deflection ratio analysis, it can be concluded that the centre model produces the most efficient structural design when compared to other curved bridge models.

Performance Comparison of Zevenet Multi Service Load Balancing with Least Connection and Round Robin Algorithm

Amikom Purwokerto University concentrates on Technology and Digital Business. This requires technology to be utilized optimally. The use of technology, especially internships, will make various jobs easier. KRS online is taking lecture schedules online via the AMIKOM Purwokerto Student website. There are several problems with the web server that arise due to the increasing need for information access, which causes the data traffic load to increase. Increased data traffic causes workload overload, resulting in server downtime. Experimental methods were used in this research to look for the causes of the web server's downtime. Then, implement the technology. The purpose is to evaluate the Zevenet load balancer performances by comparing the round-robin and least-connection algorithms. The decision is which algorithm will be used best to implement the Zevenet Load balancer to achieve a more efficient backend server traffic cluster distribution. The TIPHON standard Quality of Service parameters used in Zevenet Load Balancer performance testing are throughput, delay, jitter, packet loss, and CPU usage. The quality-of-service parameter test results show that the Zevenet Load Balancer with the round-robin algorithm has superior performance and shows less CPU usage. It is concluded that using the round-robin algorithm in implementing the Zevenet load balancer to overcome the problem of data traffic load sharing and minimize server downtime on the Student Amikom Purwokerto web server is more appropriate and more effective.

Flexural performance of full-scale two-span Nail-Laminated Timber Concrete composite slabs

This study examines the flexural performance of six 9-m full-scale two-span Nail-Laminated Timber Concrete (NLTC) composite slabs. The slabs were made with lumber beams edge-joined with double nailing, end-joined with butt joints, and the reinforced concrete topping connected with a set of notches, inclined screws, or a combination of both. The multi-span configuration of slabs reduces their deflections simply and effectively. Five-point monotonic bending tests were considered for all slabs. Before full-scale slabs, compressive and tensile pull-out tests of Timber-Concrete Composite (TCC) shear connections were performed, including notches and inclined screws. Tensile pull-out tests of shear connections were also included to emulate the negative bending moments that occur in the middle of the slabs. Failure modes, load–mid-span deflection relation, bending stiffness, and timber-concrete slip were evaluated for all slabs. A detailed 3D micro-Finite Element (FE) model of the shear connections was built in ANSYS software, whereas a macro-FE model of NLTC slabs was made in SAP2000, demonstrating a good fit for the timber-concrete interaction and the load-carrying capacity of the composite slab at the serviceability range. Moreover, an analytical elastic TCC beam with the Girhammar method was assessed and demonstrated as more precise than the traditional γ-method. Finally, an accurate prediction of the numerical and analytical (Girhammar) models for the bending stiffness at service loads up to 30% of capacity is observed, with errors in a range of 2–23% and 9–74%, respectively.

An innovative approach to a UAV tails structural design for additive manufacturing

AbstractThe innovative approach used for the design and fabrication of a UAV tails through additive manufacturing (AM) completely changes the concept of designing and manufacturing products through conventional methods. AM based processes can reduces costs, lead times, and increases design freedom, allowing parts to be tailored to specific needs. A key benefit of AM processes is the ability to produce optimized designs with reduced mass and without compromising structural effectiveness. This is possible thanks to the combination of high‐performance AM materials and the extraordinary manufacturing capability of AM technologies combined with an appropriate design‐for‐additive manufacturing (DfAM) approach, overcoming traditional manufacturing techniques. This article presents the development process of a UAV tails, based on a Design for Additive Manufacturing (DfAM) approach in order to reduce structural mass and guarantee safe operation under service loads. Therefore, by replacing the metallic component of the UAV with a techno‐polymeric one, several configurations have been developed. The optimal additive configuration achieved a weight reduction of 60% compared to a metal configuration of equivalent volume, without loss of mechanical properties. In conclusion, a feasibility study of the proposed configuration was carried out by 3D printing the vertical tail redesigned.Highlights An innovative approach by combining DfAM processes with FFF technology. Metal replacement of UAV tails with advanced technopolymers. Efficient Weight Reduction about 60% compared to metal structure. Single‐piece design simplifies assembly and manufacturing. UAV redesigned tails with enhanced efficiency, adaptability, and feasibility.

An enhanced micromechanical rock–pile interface model with application to rock‐socketed pile modeling

AbstractThe increasing use of rock‐socketed piles highlights the importance of developing a suitable design method for their bearing capacity. This study quantifies the shear behavior of the rock–pile interface, which generally dominates the bearing capacity of rock‐socketed piles under service load. A micromechanics‐based rock–pile interface model with idealized nonuniform profile is proposed with two enhancements: (1) the slip line method together with nonlinear Hoek–Brown failure criteria is integrated to identify the critical shear displacement of rock asperity; and (2) the residual stage of shear behavior is properly considered with the rounding progress of sheared rubbles. The enhanced interface model is first validated by the direct shear test results under constant normal stiffness. Then, the interface model is implemented via user‐defined FRIC into the finite element code ABAQUS without the need of explicitly building the rock–pile interface profile. Accordingly, the model is applied to simulate and analyze two field cases involving rock‐socketed piles. Comparison between the predictions and field observed results shows this method can well capture the axial load transfer behavior of pile socket into weak rock.

Viscoelastic Model of Fiber Concrete under Dynamic Loading Considering the Effect of Initial Stresses

Statement of the problem. The objective of the study was to identify the peculiarities of the stress-strain state of fiber concrete under two-stage static-dynamic loading and to develop a variant of a nonlinear viscoelastic model of concrete reinforced with steel fiber. Results. The paper presents the results of experimental tests of fiber-concrete prisms under quasi-static, dynamic and static-dynamic loading conditions. On the basis of the analysis of experimental data it is established that at dynamic loading there is an upward shift of the boundary of macrocrack formation. A nonlinear viscoelastic model of fiber concrete is proposed and compared with experimental data, taking into account the effect of the initial stress level caused by the service load. The parametric study performed using this model shows that as the loading rate increases, a more intensive dynamic hardening of concrete is observed. Conclusions. The proposed viscoelastic model of fiber concrete allows to estimate the ultimate strength and corresponding strain under static-dynamic mode of loading due to an accidental action caused by the collapse of one of the load-bearing members of the structural system of the building.

Measurement of laser shock peening induced residual stress by nanoindentation and comparison with XRD technique

Laser shock peening (LSP) is an innovative surface treatment technique successfully used to improve fatigue performance of metallic components. It is based on the application of high intensity laser and suitable overlays with the aim to generate high pressure shock waves on the surface of the mechanical part to be treated. Shock waves generate severe plastic deformations and, consequently, compressive residual stresses (RS). An accurate measurement of these latter is crucial for predicting the resistance of treated parts under service loads and to assess the effectiveness of LSP process.In this paper, a non-destructive method, based on the nanoindentation technique and finite element analysis (FEA), was developed to measure the RS generated by LSP process on AA-7050-T451 samples. In particular, the methodology is based on the analysis of the nanoindentation peak load variation generated by the presence of residual stresses on a component. Obtained results were compared, for validation, with the measurements carried out by the most consolidated X-ray diffractometer (XRD) technique. The results showed a satisfactory agreement between the two techniques, revealing nanoindentation as a promising and reliable method for characterizing RS induced by LSP.

Optimization design and 3D printing of curvilinear fiber reinforced variable stiffness composites based on polar coordinate sweeping

The 3D printing of curvilinear fiber reinforced variable stiffness structures (CFRVSSs) provides new opportunities for innovative design and fabrication of complex components in response to rising demand for high-performance and lightweight components in aerospace, railway, and other fields. In this paper, we get rid of the limitations of existing research on the structure of open-hole plates and aim to investigate complex structures with large curvature. Considering the characteristics of 3D printing and service load, an optimization design method based on polar coordinate sweeping was proposed for CFRVSSs. The consistency factors between the fiber distribution (direction, content) and stress distribution (direction, magnitude) were established. Through quantitative analysis, the directional consistency factor between the optimized local fibers and stresses was enhanced by a factor of 3.69, and the correspondence factor between the fiber content and the stress magnitude was enhanced by a factor of 3.8. Furthermore, the calculation method for curve spacing and curve arc length based on polar coordinate sweeping were proposed to minimize the forming errors caused by CFRVSSs 3D printing. The design and fabrication of a high-speed rail tie rod structure were carried out using the optimization design method and 3D printing process in current research. As a result, the ultimate tensile strength of the optimized curvilinear fiber reinforced tie rod structure increased by 156 % compared to the unoptimized straight-line fiber reinforced tie rod structure. The method proposed in this study has a wide range of applications for the design and fabrication of complex components with large curvature in the engineering field.

Comparative Study on the Use of Reinforced Concrete Piles and Rock Columns in Weak Foundation Strata in Zambia

All structures require transfer of loads to the competent strata. Where less competent soil strata is encountered, the use of either pile foundations or ground improvement methods is adopted. In Zambia, reinforced concrete pile foundations are the commonly used deep foundations since, to transmit loads to stronger strata by utilizing foundation soil skin friction, end bearing or both, while also guarding against differential settlement. Unfortunately, this has proved to be an expensive construction method and it requires the use of materials processed in high energy consuming industries. Stone columns, among many techniques, are one of the most commonly used soil improvement techniques. Although this ground improvement method is considered to be relatively cheaper than reinforced concrete piles, as it only requires the use of coarse aggregates that are widely available in Zambia, it is yet to be employed in the Zambian construction industry. As such this research aimed at comparing the use of the two construction techniques by using Finite Element Method (FEM) Analysis Software Plaxis V20 (2D and 3D) under similar loading conditions and soil properties (Lusaka West, Zambia). The resulting models and simulations were used for comparison of the various parameters which included axial loading (and lateral bulging for stone columns) for skin friction characteristics of the models. In the case of stone columns, the effects of internal angle of friction, spacing and surcharge settlement behaviour were taken into account to come up with the best performing stone column model. For calibration and validation purposes, a number of physical experiments were modelled and simulated so as to produce models that matched their predecessors as much as possible, then the soil properties were adjusted to match those obtained from a local site in Lusaka West, Zambia. Based on the results, concrete pile foundations were found to endure a larger loading before attaining the ultimate failure as compared to stone columns, but the serviceability loading state of stone columns was quite similar to that of the concrete pile foundation. Hence for structures requiring relatively low serviceability loading state on relatively large ground surface area, stone columns proved to be a reasonable construction alternative for pile foundations.

A Multi-Agent RL Algorithm for Dynamic Task Offloading in D2D-MEC Network with Energy Harvesting.

Delay-sensitive task offloading in a device-to-device assisted mobile edge computing (D2D-MEC) system with energy harvesting devices is a critical challenge due to the dynamic load level at edge nodes and the variability in harvested energy. In this paper, we propose a joint dynamic task offloading and CPU frequency control scheme for delay-sensitive tasks in a D2D-MEC system, taking into account the intricacies of multi-slot tasks, characterized by diverse processing speeds and data transmission rates. Our methodology involves meticulous modeling of task arrival and service processes using queuing systems, coupled with the strategic utilization of D2D communication to alleviate edge server load and prevent network congestion effectively. Central to our solution is the formulation of average task delay optimization as a challenging nonlinear integer programming problem, requiring intelligent decision making regarding task offloading for each generated task at active mobile devices and CPU frequency adjustments at discrete time slots. To navigate the intricate landscape of the extensive discrete action space, we design an efficient multi-agent DRL learning algorithm named MAOC, which is based on MAPPO, to minimize the average task delay by dynamically determining task-offloading decisions and CPU frequencies. MAOC operates within a centralized training with decentralized execution (CTDE) framework, empowering individual mobile devices to make decisions autonomously based on their unique system states. Experimental results demonstrate its swift convergence and operational efficiency, and it outperforms other baseline algorithms.

Imperfection Sensitivity Detection in Pultruded Columns Using Machine Learning and Synthetic Data

Experimental and theoretical solutions have shown that imperfections in wide-flanged structural columns may reduce the failure load of the column by as much as 30% with respect to that of a perfect column. Therefore, the early detection and prevention of such imperfections, which would likely reduce the load capacity of a structure, are critical for avoiding catastrophic failure. In the present article, we show how machine learning may be used to detect imperfection sensitivity in pultruded columns using observable column deformations occurring at loads as low as 30% of the design load. Abaqus simulations were used to capture the behavior of such columns of various lengths under service load. The deformations found from the simulations were used to train the machine learning algorithm. Similar deformations could be easily collected from in-service columns using inexpensive instrumentation. With over 3000 test cases, 95% accuracy in the correct detection of imperfection sensitivity was found. We anticipate that the proposed machine learning pipeline will enhance structural health monitoring, providing timely warning for potentially compromised structures.

Deep Reinforcement Learning-Based Task Offloading and Load Balancing for Vehicular Edge Computing

Vehicular edge computing (VEC) effectively reduces the computational burden on vehicles by offloading tasks from resource-constrained vehicles to edge nodes. However, non-uniformly distributed vehicles offloading a large number of tasks cause load imbalance problems among edge nodes, resulting in performance degradation. In this paper, we propose a deep reinforcement learning-based decision scheme for task offloading and load balancing with the optimization objective of minimizing the system cost considering the split offloading of tasks and the load dynamics of edge nodes. First, we model the mutual interaction between mobile vehicles and Mobile Edge Computing (MEC) servers using a Markov decision process. Second, the optimal task-offloading and resource allocation decision is obtained by utilizing the twin delayed deep deterministic policy gradient algorithm (TD3), and server load balancing is achieved through edge collaboration using a server selection algorithm based on the technique for order preference by similarity to the ideal solution (TOPSIS). Finally, we have conducted extensive simulation experiments and compared the results with several other baseline schemes. The proposed scheme can more effectively reduce the system cost and increase the system resource utilization.

A Comprehensive Review of Fatigue Strength in Pure Copper Metals (DHP, OF, ETP)

Due to their exceptional electrical and thermal conductivity properties, high-purity copper (Cu-DHP) and copper alloys of similar composition, such as electrolytic tough-pitch (ETP), oxygen-free electronic (OFE) and oxygen-free (OF), have often been used in the manufacture of essential components for the electrical, electronic and power generation industries. Since these components are subject to cyclic loads in service, they can suffer progressive structural damage that causes failure due to fatigue. The purpose of this review is to examine the most relevant aspects of mechanical fatigue in Cu-DHP, ETP, OFE and OF. The impact of many factors on fatigue strength (Se), including the frequency, temperature, chemical environment, grain size, metallurgical condition and load type, were analyzed and discussed. Stress–life (S-N) curves under zero mean stress (σm = 0) were found for high-cycle fatigue (HCF). For non-zero mean stress (σm ≠ 0), stress curves were based on a combination of Gerber, Soderberg and ASME elliptic failure criteria. Stress–life (S-N) curves were also developed to correlate fatigue strength (Se) with stress amplitude (σa), yield strength (Syp) and ultimate strength (Sut). Finally, for low-cycle fatigue (LCF), strain–life (ε-N) curves that establish a relationship between the number of cycles to failure (N) and total strain amplitude (εplastic) were determined. Hence, this review, as well as the proposed curves, provide valuable information to understand fatigue failure for these types of materials.

ОСОБЕННОСТИ ОБРАБОТКИ БОЛЬШИХ ОБЪЕМОВ ДАННЫХ ПРИ УПРАВЛЕНИИ БЕСПИЛОТНЫМИ ЛЕТАТЕЛЬНЫМИ АППАРАТАМИ НА ОСНОВЕ ИМИТАЦИОННОГО МОДЕЛИРОВАНИЯ

The paper discusses the principles of constructing a simulation model to study the influence of the type of transactions to a database management system on the reception and assimilation of data from unmanned aerial objects. The proposed study examines the interaction of unmanned aerial objects with a data receiver when implementing the monitoring process. The interaction time of the unmanned aerial vehicle with the dispatch center is modeled based on various types of transactions to the database, taking into account the differentiation of the database management system server load and their impact on the feedback time from the unmanned aerial vehicle, which will allow assessing the risks of control in an online mode.

Structural behavior of concrete slabs made of seawater sea sand and reinforced with GFRP reinforcement under flexural loading

Seawater sea sand concrete (SSC) structures reinforced with glass fibre-reinforced polymer (GFRP) reinforcing bars offer a promising alternative to conventional reinforced concrete, especially in the face of depleting clean water and river sand. However, comprehensive studies on the flexural performance and cracking behaviour of SSC structures reinforced with GFRP reinforcing bars are currently limited. This study delves into the influence of longitudinal reinforcement ratio on the flexural and cracking behaviour of SSC slabs reinforced with GFRP reinforcing bars under quasi-static loads, as well as the bond behaviour of GFRP reinforcing bar and SSC. The experimental program included 9 large-scale slabs and 30 pull-out specimens, revealing that the longitudinal reinforcement ratio significantly affects the flexural performance, particularly at higher load levels. Increased reinforcement ratio improved flexural stiffness, reduced deflection and crack width, enhancing load-carrying capacity and energy absorption. Although bond strength decreased at the serviceability load (up to 47%), a marginal reduction was observed at the ultimate state. The study proposes a semi-empirical equation for the effective moment of inertia of SSC slabs, accounting for the influence of reinforcement ratio on cracking moment and reduced bond strength of GFRP reinforcing bars. This equation aligns well with experimental results, demonstrating low error and coefficient of variation.

The Influence of Shear Connector Numbers on The Flexural Characteristics of Composite Geopolymer Reinforced Concrete Beams With Small Transverse Web Openings

Geopolymer concrete is a sustainable alternative to conventional Portland cement-based concrete, utilizes industrial by-products like fly ash and slag. Renowned for its reduced carbon footprint and exceptional durability, geopolymer concrete is gaining global recognition in eco-friendly construction practices. Steel composite reinforced geopolymer concrete beams combine the strength of steel with the sustainability of geopolymer concrete, offering a sturdy structural solution. This collaborative approach results in a composite material with excellent crack resistance and durability, harnessing the advantageous properties of both steel and geopolymer concrete. These beams present a promising option in construction by striking a balance between structural performance and environmental considerations through the incorporation of geopolymer technology. The focus of the current study is to investigate the impact of steel connectors on the flexural behavior of composite reinforced geopolymer concrete beams that have transverse web opening through experimental methods. The experimental program involves casting four geopolymer beams, with one being a traditional geopolymer-reinforced concrete beam (Control). The remaining three specimens are composite reinforced geopolymer concrete beams, each featuring a bottom 5 mm steel plate with six, twelve, and eighteen connectors near each support. The beams share common dimensions, with a total length of 1600mm, a height of 250mm, and a width of 180mm. The results showed that increasing the number of connectors near supports increased the resulted flexural behavior of beams. Such connectors increased the first crack load from 6.67% to 42.22% while the service load increased from 16.22% to 36.20%. the load carrying capacity was also increased from 12.00% to 24.80%. Finally, the mode of failure was moved from traditional tension failure to concrete crushing.

Load transfer mechanisms for capped and uncapped non-displacement piles above tunnels: a centrifuge modelling study

This paper presents results from geotechnical centrifuge tests of tunnelling in dry dense sand beneath loaded non-displacement piles, focusing on the effect of a pile cap (representative also of a raft or grade beam) in contact with the soil surface on load transfer mechanisms. Experiments included loading tests to ascertain the foundation capacity and load-displacement response in the presence/absence of an underlying model tunnel. Individual ‘reference’ pile response is compared for cases with and without a pile cap, including pile displacements and load distributions between the head, shaft, and base; the case of ‘friction’ piles with a compressible base are also considered. Results show that uncapped piles with relatively large service loads experience ‘geotechnical failure’ (i.e. large settlements or a significant increase in settlement rate with tunnel volume loss) in order to mobilise base or shaft resistance. Pile caps are shown to reduce settlements and prevent geotechnical failure of both reference and friction piles; they also improve the post-tunnelling foundation performance under loading. The paper provides evidence to enable engineers to consider the beneficial role of shallow elements in contact with the surface on the performance of pile foundations both during and subsequent to tunnelling.