Summary Structural health monitoring (SHM) systems evaluate the state of the infrastructures by analyzing the monitored responses. As measuring all target responses is difficult to accomplish due to technical or economic limitations, converting other easy‐measuring responses to the target one is a popular way. Relative approaches are separated into data‐driven and model‐driven ones. This paper proposes a deep learning‐based framework to reconstruct multitypes of full‐field responses. The adopted architecture is a convolutional neural network (CNN) with an autoencoder structure and skip connections. Varied from other data‐driven approaches, the training set in this paper is the responses computed by a finite element model (FEM), with which the CNN can learn the full‐field mapping relationships among varied response types. Therefore, the proposed framework is data‐model‐co‐driven. In the numerical simulation section, a simply‐supported beam and a continuous beam bridge have been adopted to discuss the influence of hyperparameters (training epoch, kernel size, skip connection, and bottleneck size), sensor arrangement, modeling error, and measurement noise, which indicates that the framework applies to the in‐field structures. Furtherly, a laboratory experiment has been conducted to validate the framework using a two‐span continuous bridge with obvious FEM error. All results have shown that the deep‐learning‐based response reconstruction algorithms can obtain the training set from not only in‐field measurements, but also numerical models to improve the diversity of training data.

As a structural health monitoring (SHM) system can hardly measure all the needed responses, estimating the target response from the measured responses has become an important task. Deep neural networks (NNs) have a strong nonlinear mapping ability, and they are widely used in response reconstruction works. The mapping relation among different responses is learned by a NN given a large training set. In some cases, however, especially for rare events such as earthquakes, it is difficult to obtain a large training dataset. This paper used a convolution NN to reconstruct structure response under rare events with small datasets, and the main innovations include two aspects. Firstly, we proposed a multi-end autoencoder architecture with skip connections, which compresses the parameter space, to estimate the unmeasured responses. It extracts the shared patterns in the encoder and reconstructs different types of target responses in varied branches of the decoder. Secondly, the physics-based loss function, derived from the dynamic equilibrium equation, was adopted to guide the training direction and suppress the overfitting effect. The proposed NN takes the acceleration at limited positions as input. The output is the displacement, velocity, and acceleration responses at all positions. Two numerical studies validated that the proposed framework applies to both linear and nonlinear systems. The physics-informed NN had a higher performance than the ordinary NN with small datasets, especially when the training data contained noise.

No abstract

<p>Structural health monitoring (SHM) techniques evaluate the state of the structures and detect damages based on the analyses of the monitored responses. As the measurement of all target responses can be difficult due to various limitations, reconstructing the target responses using measured data is necessary. To reconstruct the response data in the field of structural health monitoring, this paper proposes a multi-end deep convolutional network with an encoder-decoder structure and skip connections. The responses are computed by the finite element model and then divided into the training set. The proposed network model is trained to map the relationships among the various responses of involved positions. Varied measured data can be fused to reconstruct different desired responses at multi-position, leveraging a single network. Two numerical simulations are conducted to demonstrate the proposed method's applicability.</p>