Accurate prediction of the remaining useful life (RUL) of batteries is of great importance for the health management of different equipment and machines, such as electric vehicles and smartphones. It gives operators information about when the battery should be replaced. Predicting the batteries’ RUL using the data only from early cycles can also be beneficial for manufacturers. For example, it can reduce the batteries’ testing costs during the research and development phase. This article focuses on batteries’ RUL early prediction using data-driven methods. The differential capacity ( $dQ/dV)$ and differential voltage ( $dV/dQ)$ curves can reveal the potential capacity and voltage of a battery, respectively, and they are known to be indicators of the batteries’ degradation. We will present a practical method for batteries’ RUL early prediction using features extracted from those two curves. Accordingly, 19 features generated from the $dQ/dV$ and $dV/dQ$ curves are analyzed and extracted. The Sparse Bayesian Learning (SBL) method is a popular machine learning method in the field of RUL prediction, and it is used to achieve an RUL early prediction for batteries. In the end, the training and test errors are investigated to evaluate the presented method’s efficiency. Moreover, we compared our results with two other methods (lasso and elastic net), which have been recognized as best performing methods in this field so far, and the comparisons showed our proposed method outperforms those two methods in the term of accuracy. The presented method is generic and can be used for RUL early prediction of different batteries.

In many applications, it is not easy to generate enough whole-life data for training a deep neural network, which may reduce the performance of a health indicator (HI). To solve this problem, a new HI construction method based on deep transfer learning is proposed in this article. First, a new multiscale domain-adversarial neural network is proposed to extract representative features from the data collected under different working conditions. By introducing the maximum mean discrepancy regularizer and the Laplace regularizer, this model can enhance features' discriminant ability for incipient fault and exploit overall degradation information simultaneously. Then, a new HI is obtained via dimensionality reduction based on the extracted features. Moreover, a new HI assessment metric is proposed to effectively estimate the performance of the obtained HI. This new index considers not only the nonlinear correlation but also the geometric similarity of the degradation processes under different working conditions. Comparative experiments are conducted on the IEEE PHM Challenge 2012 bearing data set and the XJTU-SY bearing data set. The results show that the proposed method outperforms several typical transfer learning methods and nontransferable deep learning methods in terms of monotonicity and correlation.

One of the most significant and growing research fields in mechanical and civil engineering is structural reliability analysis (SRA). A reliable and precise SRA usually has to deal with complicated and numerically expensive problems. Artificial intelligence-based, and specifically, Deep learning-based (DL) methods, have been applied to the SRA problems to reduce the computational cost and to improve the accuracy of reliability estimation as well. This article reviews the recent advances in using DL models in SRA problems. The review includes the most common categories of DL-based methods used in SRA. More specifically, the application of supervised methods, unsupervised methods, and hybrid DL methods in SRA are explained. In this paper, the supervised methods for SRA are categorized as multi-layer perceptron, convolutional neural networks, recurrent neural networks, long short-term memory, Bidirectional LSTM and gated recurrent units. For the unsupervised methods, we have investigated methods such as generative adversarial network, autoencoders, self-organizing map, restricted Boltzmann machine, and deep belief network. We have made a comprehensive survey of these methods in SRA. Aiming towards an efficient SRA, DL-based methods applied for approximating the limit state function with first/second order reliability methods, Monte Carlo simulation (MCS), or MCS with importance sampling. Accordingly, the current paper focuses on the structure of different DL-based models and the applications of each DL method in various SRA problems. This survey helps researchers in mechanical and civil engineering, especially those who are engaged with structural and reliability analysis or dealing with quality assurance problems.

Rolling bearing health condition evaluation under time-varying operating speed requires information about shaft rotational speed. However, it is not always feasible to install a sensor such as a tachometer or an encoder to collect rotational speed signals due to space limitations or safety considerations. Time-frequency ridge estimation is an alternative approach that directly extracts the fault-related characteristic curves in a time–frequency representation (TFR) without phase reference. The cost function ridge estimation (CFRE) is the most widely used contemporary ridge estimation method. However, there is no explicit principle for the selection of search bandwidth. Moreover, the definitions of weight factor and cost function lack rationality, which causes the CFRE to degrade partially into coarse direct maximum ridge estimation (DMRE) or possibly extract the least frequency-varying curve. A time–frequency ridge estimation method, which is called adaptive CFRE (ACFRE), is developed in this study to address these deficiencies. We improve the original cost function and redefine its mathematical model to better balance searching for peak amplitude and guaranteeing a continuous curve. Numerical simulations and experimental investigations show that the proposed ACFRE is more robust to interferences derived from adjacent curves and noise than CFRE and another well-known ridge estimation approach, fast path optimization (FPO). The proposed method has a smaller root mean square error (RMSE) in estimating bearing characteristics under variable speed conditions.

In this study, active vibration control of a cantilevered flexible beam structure equipped with bonded piezoelectric sensor/actuators is investigated. The linear quadratic regulator technique together with an observer is adopted to design the controller as well as to provide the full-state feedback. Two different approaches are subsequently used for simultaneously integrated optimization of the controller and observer parameters. In the first approach, a linear experimental model of the system is obtained using identification techniques, and the optimization is then performed based on a computer simulation of the system. However, in the second approach, a hardware-in-the-loop optimization scheme is proposed and applied to automatically tune the parameters. In both cases, a hybrid metaheuristic algorithm called tabu continuous ant colony system is utilized to find the optimal controller and observer. The integrated performance is experimentally evaluated to demonstrate effectiveness of the proposed hardware-in-the-loop optimization method. Finally, a comparison is made between the frequency response of the current controller with a pole placement controller from the literature in order to assess the improvement achieved through hardware-in-the-loop optimization.