In this paper, a novel purely data-driven modal parameter identification approach for time-varying structural systems based on dynamic mode decomposition with sliding-window recursive mechanism is presented, which can address the drawbacks of parameterized modeling methods. The effectiveness of proposed method is illustrated through representative numerical simulations with time-varying dynamic behaviours, exhibiting a good engineering application prospect.

Quasi-zero stiffness isolators have a broad application prospect in effectively suppressing low-frequency vibration, but most quasi-zero stiffness isolators for low-frequency vibration isolation are only applicable to rated loads. Considering the strong nonlinearity and structural complexity of quasi-zero stiffness vibration isolators, small changes in system parameters may lead to dramatic deterioration of vibration isolation performance in an undesired direction.In this paper, the dynamics model of the vibration isolator under force excitation and displacement excitation is established, and the effects of the three uncertainties of damping, stiffness and excitation amplitude acting individually and coupled on the steady-state displacement response of the quasi-zero stiffness vibration isolator are analyzed by using the chaotic polynomial expansion method. By post-processing the coefficients of the established chaotic polynomial model, the Sobol` global sensitivity indices of each uncertainty parameter at different frequency bands under the coupling action are investigated.

In engineering, the adjustment of large hoop truss mesh antennas (LHTMA) reflector relies on manual labor. The adjustment method is usually complex, time-consuming, and often difficult to achieve high accuracy. The paper proposes an active adjustment method for LHTMA based on inertial actuators. Firstly, the driving method of inertial actuators based on piezoelectric ceramics was studied. Then, optimization methods were used to study the active adjustment algorithm for LHTMA, and genetic algorithms were used to solve the optimization problem. Finally, the feasibility of the active adjustment method was demonstrated through a simulation example of LHTMA with 8m aperture.

This paper proposes a topology optimization method for permanent magnet (PM) motors using the Gabor filter, which is widely used in image processing. In the proposed optimization, we maximize the average torque and minimize the torque ripple simultaneously. The proposed method leads to a PM motor with slit-shaped magnetic cores, which is difficult to obtain conventional approaches. It is shown that the proposed optimization results in PM motors with slit-shaped magnetic cores, which have better torque performance than a conventional model.

This study focuses on the topology optimization of the iron yoke used in Nuclear Magnetic Resonance (NMR) logging sensors for logging-while-drilling applications. A novel structure is designed to maximize the field strength and linearity of the static field for NMR logging sensors using topology optimization method based on NGnet-method.

Engineering structures subjected to stationary random loads tend to have a large impact on the stability of the structure, and the load environment during its operation must be accurately identified to ensure the safe operation of the structure. In this work, a stationary random load identification method based on ensemble empirical modal decomposition (EEMD) and deep recurrent neural network (RNN) is proposed. The method designs a deep RNN model with one gated recurrent unit (GRU) layer, one bi-directional long short-term memory (Bi-LSTM) layer, two long short-term memory (LSTM) layers and two feedforward layers in sequence, and uses decomposed load and response data as the output and input of the network. The effectiveness of the proposed method is verified by simulation data of a three-degree-of-freedom linear system and experimental data of a clamped beam.The results show that the proposed method is with higher accuracy than using only deep recurrent neural network. Further, the proposed method is also effective under undefined conditions.

This paper presents a topology optimization of a rotor core for asymmetric interior permanent magnet (IPM) motors. In the proposed optimization, the rotor structure is represented by a linear combination of normalized Gaussian functions to improve the average torque and torque ripple of an asymmetric IPM motor simultaneously. It is shown that the optimized asymmetric IPM motor, which improves the reluctance torque effectively, has higher average torque than the optimized symmetric one.

Sparse regularization methods have been extensively used in impact force identification to accurately determine the time history and locations of the impact force. However, in the context of impact forces, especially in structural health monitoring, accurate impact localization is far more critical than reconstruction time history. Furthermore, increasing the number of monitoring points to enhance force localization accuracy leads to a significant increase in the dimensionality of the transmission matrix. In this paper, a reduced iteratively reweighted l1-norm minimization algorithm (rIRI1) is proposed, which combines the reweighted lp-norm regularization with reduced control equation method. By taking into account the sparsity of impact force, the reduced transfer matrix is reconstructed to alleviate computational difficulty. Several simulations and experiments are conducted on an edge-fixed plate to compare the performance of IRL1 and rIRI1 under different underdetermined systems with varying numbers of monitoring points. The numerical and experimental results demonstrate the proposed method a substantial enhancement in computational efficiency compared to IRL1.

In order to improve the identification method of impact load using transient statistical energy analysis, which requires the known point mobility of excited subsystem to compute load, an identification method employing the basis vectors is proposed. The subsystem applied with impact load is firstly identified using the average kinetic energy of all subsystems based on the established transient statistical energy model of complex structure. On this basis, the basis vector is defined according to the distribution position, signal energy of impact load and its arrival time to sensor installed on the subsystem, so as to identify the specific position of impact load and obtain the impact signal of impact position. The time history is ultimately reconstructed by combining the identified input energy of impact load and the impact waveform. The method provides an effective way to identify the impact load of complex structures through improving the shortcomings of existing methods.

A robust optimization method, that can obtain a solution with high robustness in practical

settings with unpredictable fluctuation, for quadratic unconstrained binary optimization problems is proposed. The proposed method is applied to the optimal permanent magnet (PM) design problem of an electronic magnetic device. Even in the case of occurrence of a fluctuation of material property of the PM, it is possible to obtain a robust PM structure in which a value close to the target magnitude of magnetic flux density can be obtained.

In this paper, a fast inversion method for reconstructing the vibration loads of pipeline structures based on a dynamic transfer matrix method was proposed. The method involves a scheme to calculate the harmonic vibration responses as well as the dynamic transfer matrices of the pipeline structures with the finite element model, and an inverse algorithm to recover the unknown loads from vibration information at limited measuring points. Pipeline vibration experiments were also conducted to validate the numerical model and the proposed reconstruction method. Based on the simulated and measured vibration signals, the validity and efficiency of the proposed method were confirmed in regard to the inversion of vibration loads on pipeline structures from vibration signals at limited measuring points.

It is demanding to optimize the cable ampacity to make full use of the cluster. Since the optimization of the cable ampacity in a cable cluster is a hard constrained global optimization problem, and the final solution may be located on the constrained boundary, the existing optimization methodology will incur deficiencies in solving such a problem. In this respect, a hybrid optimizer of an improved genetic algorithm (IGA) and a local search adaptive tabu search (ATS) is proposed to optimize the ampacity in a cable cluster. The proposed optimizer uses a dual population strategy in IGA to exploit promising infeasible solutions effectively during the evolutionary process. A new neighborhood structure is proposed in ATS to generate neighborhood solutions directionally, helping local searches near constraint boundaries. The superiorities of the proposed method are confirmed by the optimization results of a test function and a cable cluster.

The number and location of nodes is crucial for the accuracy of radial basis function (RBF) meshless method. To promote the accuracy of electromagnetic numerical calculations, an adaptive refinement algorithm based on an error threshold in RBF meshless method is proposed. To identify the exact area where new nodes need to be generated, an error threshold is proposed and a hybrid node generation method is developed to promote the efficiency of node generation. A metal box example is used to test the proposed method. The results demonstrate the proposed method have the ability to improve the calculation accuracy of RBF collocation meshless method.

Our research group has developed a machine-learning method to estimate the radiation source distribution from γ-ray spectra in a space. In the previous study, we estimated radiation sources distributed in a 2D plane and achieved a high estimation rate. This paper extends the method to 3D space and discusses the estimation methods and results. In the 3D extension, it is examined that the estimation accuracy depends on the number of detection planes of the γ-ray spectrum used for estimation, and the effectiveness of estimation in 3D space is revealed.

Large-scale optimization has become the hot spot of practical engineering problems due to the complex structure of modern system. An improved fireworks algorithm (FWA) is proposed to solve the large scale optimization problem. To promote the ability of jumping out of the local optimal solution, differential sparks generation mechanism and a novel reinitialization mechanism are developed. To speed up the convergence, a dual-channel selection mechanism is proposed. CEC2013 test suite is used to test the performance of the proposed method. The numerical results demonstrate that the proposed method has better performance among the FWA variants.

“Tie” constraint is commonly used to establish the link relation of structural finite element model (FEM) in engineering. In this paper, the confined area is taken as the updated object, and modal test data is used to update the structural finite element model. Artificial neural network (ANN) is used to describe the nonlinear relationship between the natural frequency and the confined area of the structure. Training the network with FEM calculation results, and updating the FEM by test natural frequency. The results show that the error of natural frequency between the modal test and modal analysis of FEM is small, so the model has high precision.

As the modern antenna developing towards systematization and multi-function, antenna optimizations are often involved with large scale of variables. Fireworks algorithm (FWA) based on differential grouping is proposed to solve these high dimensional antenna designs. To search the variable space efficiently in a vast variable space, differential grouping method is introduced to identify the dependent variables and divide the optimization into sub-domains. To cooperate the evolutions in variable sub-domains, an information exchange and a monitor strategy is developed. An UWB antenna is proposed test the effectiveness and efficiency of the proposed method.

A novel data-DRNN method is proposed for time-domain load identification of a cylindrical structure subjected to random base excitation of electromagnetic shaker. The data-DRNN model comprises two Long Short-Term Memory (LSTM) layers and one Bidirectional LSTM (BLSTM) layer, trained using a large dataset of measured acceleration. The effectiveness and accuracy of the proposed method are validated under various temperature conditions using quartz lamp heater. Furthermore, the model's generalization capability is evaluated by different Power Spectral Density (PSD) target spectrums of excitation. The results indicate that data-DRNN has great accuracy and generalization ability, making it a promising choice for load identification of complex structures.

Research on the use of limited sensor measurement information to identify modal parameters in underdetermined situation has important engineering application values. This paper proposes an underdetermined modal parameters identification method based on block term decomposition. By transforming the underdetermined modal parameters identification problem into tensor uniqueness decomposition problem, the connection between modal parameters identification and tensor decomposition is established, and the algorithm is optimized to solve the problem of uncertain sequence of source signals after decomposition. Finally, through a numerical simulation example and an experiment of a certain type of solid rocket motor structure, the ability of the proposed method in dealing with the identification of underdetermined modal parameters is verified, demonstrating good engineering application prospects.