A modal transfer enhanced deep learning for structural dynamic response with sparse spatial data

Estimating population size from spatially discrete sampling data is a routine task of ecological monitoring. This task may however become challenging in the case that the spatial data are sparse. The latter often happens in nationwide pest monitoring programs where the number of samples per field or area can be reduced to just a few due to resource limitation and other reasons. In this rather typical situation, the standard (statistical) approaches may become unreliable. Here we consider an alternative approach to evaluate the population size from sparse spatial data. Specifically, we consider numerical integration of the population density over a coarse grid, i.e. a grid where the asymptotical estimates of numerical integration accuracy do not apply because the number of nodes is not large enough. We first show that the species diffusivity is a controlling parameter that directly affects the complexity of the density distribution. We then obtain the conditions on the grid step size (i.e. the distance between two neighboring samples) allowing for the integration with a given accuracy at different diffusion rates. We consider how the accuracy of the population size estimate may change if the sampling positions are spaced non-uniformly. Finally, we discuss the implications of our findings for pest monitoring and control.

Assimilating field data is essential to calibrate rock properties in hydrogeological applications. While such workflows are challenging for physics‐based reservoir simulators, data‐driven Artificial Neural Networks (ANN) offer a potential solution by effectively managing noisy data to infer fluid pressure diffusion p ( x , t ). Nevertheless, the sparse observation wells in hydrogeology projects make training ANN models difficult. Therefore, we employ Physics‐Informed Neural Networks (PINN) to bridge this gap between sparse spatial data. We train both ANN and PINN on a synthetic p ( x , t ) data set within a low‐permeability porous medium containing a highly permeable material. Having an identical architecture, both models integrate direct observations of the target variable, while the PINN further optimizes the diffusion equation residual. By comparing the model output as a function of the number of observation wells, we find that the PINN outperforms the ANN when spatial observations are limited to a small number of wells. Adding noise to the training data further underscores the robustness of PINN against random measurement errors or ambient noise. Furthermore, we evaluate the effectiveness of PINN in representing a high‐permeability zone without observations inside the domain. The findings suggest that accuracy is maintained even with reduced thickness of the equivalent material, although this leads to an increase in computation time. Given the sparsity of observation wells in hydrogeological applications, PINNs potentially provide a more suitable machine learning workflow than purely data‐driven ANN for reservoir modeling.

Numerical analysis of dynamic responses of laminated glass window subjected to gas explosions

It is pointed out that the damping matrix deduced by active members in the finite element vibration equation of a truss adaptive structure generally can not be decoupled, which leads to the difficulty in the process of modal analysis by classical superposition method. This paper focuses on the computational method of the dynamic response for truss adaptive structures. Firstly, a new technique of state vector approach is applied to study the dynamic response of truss adaptive structures. It can make the coefficient matrix of first derivative of state vector a symmetric positive definite matrix, and particularly a diagonal matrix provided that mass matrix is derived by lumped method, so the coefficient matrix of the first derivative of state vector can be exactly decomposed by CHOLESKY method. In this case, the proposed technique not only improves the calculation accuracy, but also saves the computing time. Based on the procedure mentioned above, the mathematical formulation for the system response of truss adaptive structures is systematically derived in theory. Thirdly, by using FORTRAN language, a program system for computing dynamic response of truss adaptive structures is developed. Fourthly, a typical 18 bar space truss adaptive structure has been chosen as test numerical examples to show the feasibility and effectiveness of the proposed method. Finally, some good suggestions, such as how to choose complex mode shapes practically in determining the dynamic response are also given. The new approach can be extended to calculate the dynamic response of general adaptive structures.

The classical step-by-step integration method is commonly used for solving structural dynamic responses, but its performance is highly sensitive to the choice of time step. A large time step can significantly reduce accuracy and stability, while a very small time step greatly diminishes computational efficiency. Data-driven modeling and analysis methods are particularly favored for their potential to replace traditional structural analysis methods. However, traditional data-driven machine learning approaches require substantial amounts of training data to predict structural dynamic responses. Moreover, these methods often function as black boxes, lacking physical interpretability, and acquiring the necessary data can be challenging in practice. To address these limitations, this paper employs a physics-informed neural network (PINN) method for structural dynamic response analysis. The PINN approach incorporates the governing equations of structural dynamics into the neural network’s loss function as physical constraints, endowing the model with physical interpretability. The feasibility and effectiveness of this method are demonstrated through case studies on structural dynamic responses under different excitation scenarios. Numerical results show that the PINN method can accurately analyze structural dynamic responses without relying on system response data, a key advantage over traditional data-driven models. Moreover, unlike the classical step-by-step integration method, the PINN approach operates as a spatial domain optimization process and is independent of time step. Given the critical role of hyperparameters in the performance of the PINN model, this paper systematically investigates the influence of various hyperparameter settings on accuracy and efficiency, uncovering patterns of their impact on structural response analysis. Replication code for this paper will be available on GitHub:https://github.com/duke-iem/AIStructDynSolve.

Recognizing human activity from highly sparse body sensor data is becoming an important problem in Internet of Medical Things (IoMT) industry. As IoMT currently uses batteryless or passive wearable body sensors for activity recognition, the data signals of these sensors have a high sparsity, which means that the time intervals of sensors’ readings are irregular and the number of sensors’ readings in time unit is frequently limited. Therefore, learning activity recognition models from temporally sparse data is challenging. Traditional machine learning techniques cannot be applicable in this scenario due to their focus on a single sensing modality requiring regular sampling rate. In this work, we propose an effective end-to-end deep neural network (DNN) model to recognize human activities from temporally sparse data signals of passive wearable sensors that improve the accuracy rate of human activity recognition. A dropout technique is used in the developed model to deal with sparsity problem and avoiding overfitting problem. In addition, optimization of the proposed DNN model was performed by evaluating a different number of hidden layers. Various experiments were conducted on a public clinical room dataset of sparse data signals to compare the performance of the proposed DNN model with the conventional and other deep learning approaches. The experimental results demonstrate that the proposed DNN model outperforms the existing state-of-the-art methods in terms of lower inference delay and activity recognition accuracy.

The success of deep learning models in fluid dynamics applications will depend on their ability to handle sparse and noisy data accurately. This paper concerns the development of a deep learning model for reconstructing turbulent flow images from low-resolution counterparts encompassing noise. The flow is incompressible through a symmetric, sudden expansion featuring bifurcation, instabilities, and turbulence. The deep learning model is based on convolutional neural networks, in a high-performance, lightweight architecture. The training is performed by finding correlations between high- and low-resolution two-dimensional images. The study also investigates how to remove noise from flow images after training the model with high-resolution and noisy images. In such flow images, the turbulent velocity field is represented by significant color variations. The model's peak signal-to-noise ratio is 45, one of the largest achieved for such problems. Fine-grained resolution can be achieved using sparse data at a fraction of the time required by large-eddy and direct numerical simulation methods. Considering its accuracy and lightweight architecture, the proposed model provides an alternative when repetitive experiments are complex and only a small amount of noisy data is available.

The dynamic response of structures is examined under a general condition of loading. It is shown that the inelastic response of structures depends on the interaction between forces and displacements existing at a section during yielding. A theory of yielding is developed in terms of forces and displacements incorporating the effects of such interactions. Based on this theory, a force-displacement relationship is derived under a general condition of loading. The use of this relationship to study the response of structures is discussed and equations of motion are derived for a simple frame subjected to simultaneous base excitation along its principal directions. To study the inelastic response of structures, under a general condition of loading, it is necessary to derive the equation of the yield surface in terms of forces acting at a section. For the special case of bending about the principal axes of a section, equations of yield surfaces are derived for various structural sections. The response of a simple frame, subjected to sinusoidal base excitation, is obtained for elastic behavior, elasto-plastic behavior and elasto-plastic behavior with interaction. The response for these behaviors is compared and it is shown that interaction causes significant changes in the response. The response of the frame is also investigated for earthquake type excitation and a series of curves are presented to show the effect of interaction on various response parameters. Use of these curves for inelastic design of structures is indicated and the implications of the effects of interaction are examined.

Prediction of the dynamic response of a slender marine structure under an irregular ocean wave using the NARX-based quadratic Volterra series

System identification of dynamic structure by the multi-branch BPNN

Li, L.; Shi, P.; Du, X., and Jiao, H., 2017. Using numerical simulation to determine the seismic response of coastal underground structures in saturated soil deposits. This paper discusses numerical simulation of seismic response of underground structures in saturated soil deposits. A fully fluid–solid coupling model is developed to simulate the dynamic behavior of saturated soils during an earthquake and is implemented into the commercial software ABAQUS with a user-defined element. The model is then applied to simulate seismic response of a structure buried in a saturated soil deposit. The effect of burial depth on the dynamic response of the underground structure is explored. The numerical simulations show that the user-defined element developed in this paper is capable of simulating the dynamic response of the underground structure in a saturated soil deposit. With the increase of burial depth, stress in the underground structure is transferred from the exterior walls, top slab, and bottom sl...

Modelling of jack-up response for fatigue under simulated service conditions

This paper investigates the role of material microstructures in structural analysis and establishes the need for microstructure-integrated constitutive models in predicting structural response. A focus is on microstructure and temperature dependency of stresses and plastic strains in a structural panel under realistic loading conditions. Structural analysis is conducted using the recently developed uncertainty-quantified parametrically homogenized constitutive model (UQ-PHCM) for near-alpha Titanium alloys like Ti6242S. PHCMs exhibit explicit microstructural dependency and are developed from homogenization of crystal plasticity finite element simulation results with machine learning. Uncertainties due to model reduction, data sparsity and microstructural variability are accounted for in the model. Structural response with the UQ-PHCMs is compared to those predicted by isotropic elasticity and $${J}_2$$ plasticity models without explicit microstructure information. Parametric studies illustrate how different uncertainties in the UQ-PHCM framework propagate to the structural response variables. The results also show the relative contribution of different microstructural features to the propagated uncertainty in structural response variables. The studies establish the UQ-PHCM as an effective tool for reliable structural analysis with consequences in material design.

The magnetorheological (MR) damping material is a kind of smart shock absorption material, and it can be made into MR dampers for reducing the vibration or dynamic response of structures. During the vibration mitigation control of structures with MR dampers, the displacement and acceleration responses are always concerned firstly because the displacement responses determine the safety of structures and at the same time the acceleration responses determine the comfort level of the human body staying in structures. That means, the control currents choice of MR dampers during vibration process is a multi-objective optimization control problem. In this paper, the particle swarm optimization (PSO) algorithm is developed to control the displacement and acceleration responses simultaneously. Numerical analysis for a five-floor structure with one MR damper installed on each floor is carried out. Simulation results of the PSO control structure are compared with those of passive control (including the On-control and the OFF-control) structures and the uncontrolled structure. Analysis results demonstrate that the PSO algorithm can reduce the displacement responses of the structure obviously, at the same time, it can reduce the acceleration responses of the structure to a certain extent. Furthermore, the PSO algorithm reduces seismic responses of structures more effectively than those of passive control structures.

Direct identification of structural parameters from dynamic responses with neural networks