查看更多>>摘要：In actual operating conditions, rolling bearings vibration signals are easily covered by heavy noise, increasing the difficulty of fault diagnosis. A fault diagnosis method based on auto regressive moving average (ARMA) model and multipoint optimal minimum entropy deconvolution adjusted (MOMEDA) algorithm is proposed to address this issue. Firstly, ARMA model is used to remove the intrinsic components and pre-whitening the signals. Then parameters of MOMEDA are optimized by Sparrow Search Algorithm (SSA), the periodic fault signals are recovered by the optimized MOMEDA and the secondary noise reduction of the signals is realized. Finally, a class of time-domain average dimensionless features, namely average pulse factor, average kurtosis factor and average margin factor, are proposed and combined with the Gini index as fault diagnosis indexes then input into ELM classifier to identify fault types. Experimental results show the proposed method can identify fault types effectively and achieve accurate diagnosis of rolling bearings.
查看更多>>摘要：Additive manufacturing processes using the direct energy deposition method, one of which is the Wire Arc Additive Manufacturing (WAAM), have gained much attention in the scientific community over the last decade. The application of WAAM to Invar, an iron-nickel and manganese alloy, with a low amount of chromium carbon, also called FeNi36 or Nivarox has been the subject of various reports due to its challenging nature. This paper utilizes and unifies research material previously investigated in this technology, taking a new approach based on the study of symmetrical phenomena that guarantee the quality of the process. On the one hand, a method of analysis of the geometry of the manufactured wall is presented based on its symmetrical quality which guarantees the maximum use of material and, on the other hand, the monitoring of the symmetry of the melting pool utilizing thermography techniques.
查看更多>>摘要：Pseudo bubble occlusion is a challenging situation for an image analysis tool as it involves a 2D projection of a 3D object. In this work, a robust algorithm has been developed to detect the bubbles from pseudo occlusion. Images of the flow are captured using a high-speed camera. The algorithm employed for bubble detection comprises Adaptive Threshold-based Image Segmentation which comprises segmentation, edge detection, segment grouping, and ellipse fitting. The bubble size is detected and the bubble size distributions for different air flow rates are obtained. The regime identification has been done based on the cell averaged projected void fraction, bubble dimensions, and the distribution of the bubbles. As the superficial air velocity increases in the range of 0-2 cm/s, bubble break up becomes prominent, and hence, the number of smaller size bubbles increases.
查看更多>>摘要：Coordinate Measuring Machines (CMM) are widely used in form inspection of free-form surfaces. Generally, the form error at each measured point is estimated using the widely known and accurate point-inversion method. This method has relatively high time complexity and cannot be preferred for fast inspection. Hence in this work, an alternative two-stage methodology based on the concept of the Voronoi diagram is proposed. In the first stage, the poles data is extracted from the Voronoi diagram of the discretized surface. In the second stage, the formerror-estimate algorithm executing in O(m log n) time estimates the errors using the poles data and the discretized surface. Numerical and experimental implementations are executed using NURBS surfaces. The proposed method's accuracy is on par with the point-inversion method and is 94.97% faster than the latter. Hence this method can be used for fast and accurate CMM and CNC based (in-situ) free-form surface inspection.
查看更多>>摘要：The purpose of this paper is to develop a rapid detector for the battery state-of-health (SOH) in field applications. The research focuses on the detection principle and implementation technology of the instrument, which differs from machine learning methods based on data mining and equivalent-circuit model methods based on state space modeling and parameter estimation. The charge transfer factor and lithium-ion diffusion factor are introduced to represent the battery SOH in the active material and lithium-ion inventory inside the battery respectively. The relationship between the two indicators and battery impedance is established, which is independent of SOC. Two indicators are obtained by measuring the charging current at a particular single frequency point within seconds. The charge current, which comprises a fixed-amplitude DC current and variant AC current is employed to provide a unified comparison base and shortens the measurement time. The rapid detector is implemented on a microcontrol unit with HRPWM technology.
查看更多>>摘要：Cutting tool plays a critical role in modern manufacturing system and tool wear prediction has a great effect on product cost and quality. Many efforts have been devoted to developing tool wear prediction models. The previous studies mainly focus on utilizing a single model during the whole prediction process named global model and thus ignore the local characteristics of different wear stages. However, paying no attention to the impact of wear stage division may lead to mismatch the distribution characteristics of datasets and overlapping of feature information, failing to get the desired accuracy. To achieve more accurate prediction results, this paper proposes a wear stage division-based tool wear prediction method (WSDTWP) based on the improved symmetrized dot pattern (ISDP) and multi-covariance Gaussian process regression (MCGPR). Firstly, according to the varying trend of tool wear value, tool wear process is divided into three stages, including initial wear stage, moderate wear stage and severe wear stage. Then, SDP technique reconstructs the original signals visually and the main SDP parameters are adaptively selected by developing an evaluation model. Further, to deal with the issue of high-dimension and small-size datasets in initial and severe wear stages, MCGPR is developed and hyperparameters are optimized by particle swarm optimization for accurately predicting tool wear. According to the varying prediction requirements, different approaches are assigned into different wear stages to achieve better performance. Finally, three cutting tests are conducted to validate the effectiveness of the proposed approach. The experimental results indicate that WSDTWP is more accurate than existing methods, which provides new theoretical and practical support for identifying tool working condition and predicting tool wear.
查看更多>>摘要：Conveyor belt tearing real-time detection is essential for industrial transportation under multi-source interference environment. Here, a deep learning-based visual detection method named YOLOv4-BELT was proposed. The multi-condition belt tearing images dataset (MBTID) is firstly produced. Afterwards, the MBTID is pre-processed by the improved Cutmix algorithm for data augmentation, which aims to enrich image background and reduce over-fitting. Next, the deep convolutional neural network CSPDarknet53 is employed for multi-scale tear features extracting and fusion, which can effectively improve the recognition capability towards complex samples. Moreover, the training performance is significantly enhanced via a proper designed multi-stage transfer training strategy. Ultimately, the previous deep-level tear features are further utilized to classification and localization tasks. The results show that the precision, accuracy, recall and F1 score of YOLOv4-BELT are 96.6%, 99.1%, 98.1% and 97.4% respectively. The detection speed reaches 21.1FPS, which significantly improves the detection accuracy and robustness compared with the state-of-the-art methods.
查看更多>>摘要：The point cloud spatial expansion (PCSE) method creates an alternative form of representing the shape of symmetrical objects and introduces additional descriptive geometric parameters. An important element of the procedure is determining the course of the axis of symmetry of cylindrical objects based on cross-sections of point clouds. Outliers occurring in laser measurements are of great importance in this case. In this study, six robust estimation methods were used to determine the coordinates of the section centers. Accuracy analysis was performed both for data simulated with the Monte Carlo method and the real data. The study showed the advantage of robust methods for the PCSE method over the classical method of least squares estimation.
查看更多>>摘要：Color deviation of the sunglass lens brings many problems to the pairing of sunglasses. In order to accurately classify the sunglass lens by color depth, a data acquisition system based on spectral analysis method is developed, which is composed of reflection integrating sphere, optical fiber spectrometer and optical fiber. Besides, the classification algorithm based on Principal Component Analysis, with Particle Swarm Optimization and Extreme Learning Machine is proposed. In which, PCA reduces the dimensions of the spectral reflectance data, PSO optimizes the input weights and hidden layer bias values of ELM, and the optimized ELM obtains a satisfactory classification through certain learning and training. This algorithm avoids the lengthy formula calculations in the traditional color classification method, and requires fewer hidden layer neurons to achieve high and stable classification accuracy in ELM. The classification accuracy of PCA-PSO-ELM and PCA-ELM, LM-BP, LSSVM is compared by the experiments. It is proved that the adoption of the proposed PCA-PSO-ELM in the color classification of sunglass lenses is feasible and effective.
查看更多>>摘要：To overcome the low efficiency of crack depth detection of steel, we explored for the first time the method based on hyper-parameters search in the field of defect depth classification. And the effect of different defect depths on the heat transfer to the metal surface during heating and cooling process was analyzed. Moreover, we de-noise the infrared thermal images by median filtering algorithm. Then we propose two time-series temperature features: the crossing temperature feature and the temperature difference feature, and compared their robustness. We perform hyper-parameter search by grid search and random search, for KNN, SVM and random forest. Experiments prove that the temperature difference feature is effective in this study. The KNN based on grid search can achieve 100% accuracy. The SVM has the highest classification efficiency, that based on grid search and random search can achieve 100% classification accuracy in 0.63 s and 0.78 s, respectively.
NSTL
Elsevier