Search

Remaining useful life (RUL) prediction can effectively avoid unexpected mechanical breakdowns, thus improving operational reliability. However, the distribution discrepancy caused by different working conditions may lead to deterioration in the prognostic task of machinery. Inspired by the idea of transfer learning, a novel intelligent approach based on dynamic domain adaptation (DDA) is proposed for the machinery RUL prediction of multiple working conditions in this paper. At first, reverse validation technology is utilized to select appropriate source samples to construct the training dataset. Then two dynamic domain adaptation networks are trained to extract domain invariant degradation feature and predict RUL, namely dynamic distribution adaptation network and dynamic adversarial adaptation network. In the dynamic domain adaptation network, the fuzzy set theory is employed to calculate conditional distribution discrepancy loss, and the dynamic adaptive factor is introduced to dynamically adjust the distribution weights. Finally, the proposed method is proved to be effective through two run-to-failure bearing datasets. Related experimental results indicate that, compared with other related RUL prediction methods, the DDA-based prognostic method not only achieves better prediction performance, but also avoids the influence of negative transfer and distribution weight fluctuation.

The internal crack defect is prone to appear in the inner part of casting products during the production process due to the influence of casting craft and on-site environment. In order to ensure the internal quality of the casting products, nondestructive examination and defect detection technology should be used to detect the internal crack of the casting product. The existing defect detection technologies have some problems such as poor generalization and low accuracy. Therefore, an internal crack defect detection method based on the Relief algorithm and Adaboost-SVM is proposed in this paper. Firstly, casting image is preprocessed by grayscale transformation, bilateral filtering, and adaptive image segmentation. Secondly, HOG feature, invariant moment feature, and LBP feature are extracted, and sensitive feature set is selected by Relief algorithm. Finally, the Adaboost-SVM is used to construct the internal crack detection model to realize the crack detection with high generalization and accuracy. The effectiveness of the method is verified by the casting image dataset collected in the actual industrial field. The experimental result reveals that the proposed method could not only extract sensitive feature set but also has better classification performance and generalization ability than other common classifiers.

Ice accretion on wind turbine blades is one of the major faults affecting the operational safety and power generation efficiency of wind turbines. Current icing detection methods are based on either meteorological observing system or extra condition monitoring system. Compared with current methods, icing detection using the intrinsic supervisory control and data acquisition (SCADA) data of wind turbines has plenty of potential advantages, such as low cost, high stability, and early icing detection ability. However, there have not been deep investigations in this field at present. In this paper, a novel intelligent wind turbine blade icing detection method based on the wind turbine SCADA data is proposed. This method consists of three processes: SCADA data preprocessing, automatic feature extraction, and ensemble icing detection model construction. Specifically, deep autoencoders network is employed to learn multilevel fault features from the complex SCADA data adaptively. And the ensemble technique is utilized to make full use of all the extracted features from different hidden layers of the deep autoencoders network to build the ensemble icing detection model. The effectiveness of the proposed method is validated using the data collected from actual wind farms. The experimental results reveal that the proposed method is able to not only adaptively extract valuable fault features from the complex SCADA data, but also obtains higher detection accuracy and generalization capability compared with conventional machine learning models and individual deep learning model.

Feature extraction is a major challenge in operational reliability assessment, which requires techniques and prior knowledge. Deep auto-encoder (DAE) is a popular deep learning method and is widely used in feature extraction. However, low generalization ability and structure parameters design are still the major problems of DAE for operational reliability assessment. To overcome the two problems, an ensemble DAE is proposed for operational reliability assessment. Firstly, different structure parameters are employed to design a series of DAEs for feature learning from the measured data. Secondly, a feature ensemble strategy is designed to enhance the generalization ability of the DAE model, in which the features learned by different DAEs are clustered to remove the irrelevant DAEs and select the more general feature subset. Finally, the operational reliability indicator is defined by the Euclidean distance of the selected features and the operational reliability model is developed. The proposed method is utilized to analyze the experimental bearings and the results indicate that the proposed method is effective for operational reliability assessment.

This study explored a new nonparametric analytical method for identifying heterogeneous segments in time-series data for data-abundant processes. A sample entropy (SampEn) algorithm often used in signal processing and information theory can also be used in a time series or a signal stream, but the original SampEn is only capable of quantifying process variation changes. The proposed algorithm, the adjusted sample entropy (AdSEn), is capable of identifying process mean shifts, variance changes, or mixture of both. A simulation study showed that the proposed method is capable of identifying heterogeneous segments in a time series. Once segments of change points are identified, any existing change-point algorithms can be used to precisely identify exact locations of potential change points. The proposed method is especially applicable for long time series with many change points. Properties of the proposed AdSEn are provided to demonstrate the algorithm’s multi-scale capability. A table of critical values is also provided to help users accurately interpret entropy results.

An improved time-varying mesh stiffness (TVMS) model of a helical gear pair is proposed, in which the total mesh stiffness contains not only the common transverse tooth bending stiffness, transverse tooth shear stiffness, transverse tooth radial compressive stiffness, transverse gear foundation stiffness and Hertzian contact stiffness, but also the axial tooth bending stiffness, axial tooth torsional stiffness and axial gear foundation stiffness proposed in this paper. In addition, a rapid TVMS calculation method is proposed. Considering each stiffness component, the TVMS can be calculated by the integration along the tooth width direction. Then, three cases are applied to validate the developed model. The results demonstrate that the proposed analytical method is accurate, effective and efficient for helical gear pairs and the axial mesh stiffness should be taken into consideration in the TVMS of a helical gear pair. Finally, influences of the helix angle on TVMS are studied. The results show that the improved TVMS model is effective for any helix angle and the traditional TVMS model is only effective under a small helix angle.

As a new type of transmission, the single-roller enveloping hourglass worm gear (SEHWG) is widely used in precision mechanical transmission systems. Time-varying mesh stiffness (TVMS) is one of the most important excitations, which has great influences on the dynamic characteristics of SEHWG systems. In this paper, a model for calculating the TVMS of the SEHWG is proposed, in which bending stiffness, shear stiffness, radial compression stiffness, and foundation stiffness are considered in the worm stiffness calculation. And bending stiffness, shear stiffness, and foundation stiffness are considered in the worm wheel stiffness calculation. The process of calculating the periodical TVMS of the SEHWG is also presented. Then, the TVMS of the SEHWG by the proposed method is compared with that of the finite element method for the model validation. Finally, the influences of the roller radii on TVMS are analyzed. The results show that the mesh stiffness of the TVMS has a tendency to rise first and then fall with the increasing roller radii.

Transfer learning has good ability to transfer knowledge for fault diagnosis under different working condition, while domain mismatches or domain shift can still occur during single-source domain transfer fault diagnosis. To alleviate the problem, a multi-source ensemble domain adaptation method is proposed for rotary machinery fault diagnosis. Firstly, multi-source and target domain anchor adapters are constructed based on class-central samples from multi-source domain. Secondly, multi-source ensemble domain adaptation transfer fault diagnosis model considering the mutual difference between multi-source domain is established to obtain multiple classifiers and prediction results. Then the classifiers with good performance are integrated to achieve final diagnosis model and results by ensemble of anchor adapters. Finally, the performance of the proposed method is verified by two experiments. The results show that the proposed method has ability to learn more comprehensive and general domain invariant diagnosis knowledge, significant diagnosis performance and robustness than other transfer learning methods.

RUL prediction of bearings plays an essential role in avoiding unwanted downtime and improving machines' reliability. A linear reliability indicator approach for RUL prediction is proposed. Multiple deep auto-encoder models are established to extract deep features, and a clustering method is employed to select the original feature set. Monotonicity is set as a criterion for evaluating and selecting original optimal features from the original feature set and a linear reliability indicator is established through it. Based on linear reliability indicator and feature interpolation, the original feature set is transformed into a mapping feature set. Finally, the mapping feature set is trained by a reliability evaluation model and the particle filter is employed to predict RUL. To demonstrate the excellent capability of the proposed approach, the vibration signal from the public PRONOSTIA bearing datasets is used. Experiment results show that the proposed approach can achieve high accuracy in RUL prediction.

Big data related to manufacturing applications has the traits such as great quantity, multi-sources, low value density, high complexity, and dynamic state. Traditional feature extraction methods are incapable of meeting real-time demands. Therefore, a robust incremental on-line feature extraction method based on PCA (Principal Component Analysis), RIPCA (Robust Incremental Principal Component Analysis), is proposed. RIPCA adopts a sliding window to update new coming data stream and to filter outliers. The proposed method could ensure the accuracy of data analysis and meet real-time demands of big data processing for manufacturing applications. A test data set based on a semiconductor manufacturing process containing 1567 records with 590 features is used to demonstrate the availability of the proposed method. Experimental results show that the method can effectively extract features of the data stream in real time with high accuracy.

Remaining useful life (RUL) prediction has been a hotspot topic, which is useful to avoid unexpected breakdowns and improve reliability. Different bearing failure behaviors caused by multiple failure modes may lead to inconsistent feature distribution, which affects the prediction model performance. To accurately predict the RUL of bearing under different failure behaviors, a transferable convolutional neural network (TCNN) is proposed to learn domain invariant features. In the proposed method, a convolutional neural network is employed to extract the degradation features. Then multiple-kernel maximum mean discrepancies are integrated into optimization objective to reduce distribution discrepancy. The trained TCNN can be used to predict RUL by feeding data. Its effectiveness is verified by a run-to-failure bearing dataset. The comparison results reveal that the proposed method avoids the influence of kernel selection, improves the performance of domain adaptation effectively, and achieves a better RUL prediction performance.

At present the majority of the research works on machining chatter detection are based on energy level monitoring of certain chatter frequencies which are premised to be invariant. However, chatter frequencies are very complicated with multi-frequency/frequency band, time-variant characteristics and are also affected by many factors in the machining process. These characteristics make the traditional chatter detection methods vulnerable under the practical complex machining conditions. In this study, the complexity of grinding chatter frequency was investigated firstly. The influence from change of machining parameters and machine tool dynamic characteristics on grinding chatter frequencies were analyzed theoretically. Then a high accuracy and robust chatter detection method adaptable for dealing with complex machining conditions was proposed. Specifically, information fusion of feed motor current and vibration signal based on cross wavelet transform was utilized to extract chatter information adaptively and completely, whereas the harmonics of forced vibration and other noises were eliminated. To assess the grinding stability, two chatter indicators, i.e. normalized spectral entropy and logarithmic spectrum distance based on the cross wavelet transform spectrogram were devised. The efficacy of the proposed method was successfully verified through grinding operations in which both constant and variable machining were carried out. The results show that the developed method is capable of robustly detecting chatter in all the tested machining conditions and both the indicators have suitable properties to enable them for adaptive chatter detection. By comparison, spectrum distance has a monotonic relationship with the severity of chatter vibration, which makes it a comparatively better indicator.

A multi-ensemble method is proposed based on deep auto-encoder (DAE) for fault diagnosis of rolling bearings. At first, several DAEs with different activation functions are trained to obtain different types of features, which are merged into a feature pool. Then, the features in the feature pool are evaluated and selected. The classifiers are constructed for each feature. The classification accuracy is used as evaluation index and good features are selected. Finally, the train data is cross-divided and multi sample-sets with selected features are constructed. Each sample-set is used to train a classifier, the final diagnosis result is obtained by majority voting among these classification results. The proposed method is validated by two experimental bearing vibrations signals and also compared with other methods. The results revealed that different types of features can be obtained by DAEs with different activation functions. The proposed method has a high accuracy and good generalization ability.

A novel self-adaptive immune genetic algorithm (SAIGA)–dynamic back propagation neural network (DBPNN) model was developed to solve the difficulty of making maximum completion time prediction (makespan prediction). By analyzing history data of makespan and its related factors in production life cycle, a prediction model based on back propagation (BP) neural network was established, weight values and threshold values of the BP neural network model were improved dynamically, and the DBPNN model was further optimized via SAIGA, thus obtaining the SAIGA-DBPNN model. The proposed SAIGA-DBPNN model was applied to an aviation enterprise’s makespan prediction. The application case suggested that the novel prediction method can yield accurate results. The usefulness of accurate makespan prediction in production life cycle as a tool for improving production efficiency was highlighted.

Large and complex equipment reliability evaluation is extremely dependent on equipment reliability experiment data, maintenance records, and failure data. With the informationalization and intellectualization of equipment (such as CNC machine tools, shield machines, and weaponry), large amounts of data (big data) will be produced during the equipment's operation. Abundant data provide a strong support for equipment operational reliability analysis in the industrial big data age, but also pose a huge challenge for reliability analysis. This paper first explores the opportunities provided by big data to promote the reliability analysis and assessment of complex equipment. Then, we mainly focus on the remaining challenges of equipment operational reliability assessment using the industrial big data method, such as the fact that most of the data reflect an intermediate state (incomplete failure state) of the equipment. We also consider a way to analyze the multiple-states of the equipment operation and correlate the multiple failure modes of the equipment operation using the big data. Moreover, a big data analysis method for calculating the reliability and predicting the residual life of gradual systems is discussed, along with a method for combining the traditional reliability calculation theory with the big data theory. All of these issues provide a significant challenge for the reliability analysis of complex equipment in the big data age.

As manufacturing systems become increasingly complex, manufacturing enterprises face the challenging need for precise, effective process capability analyses. However, acquisition of process data, characterized by larger volume and complexity, is much easier, consequently making process capability analysis more difficult. Traditional process capability analysis methods such as Cp or Cpk are not adequate for large volume process data collected over time because these methods assume that process distribution remains unchanged. Therefore, the goal of this paper is to explore the use of sample entropy (SampEn) as it relates to univariate process capability analysis. The proposed method, which alleviates the fixed distribution assumption, can identify changing process variations over time. We proposed a novel method based on Adjusted Sample Entropy (AdSEn) to quantify process variation changes. A study based on simulation data sets showed that the proposed method provides adequate process capability information.Copyright © 2015 by ASME

The goal of this paper is to shorten the gap of industrial virtual simulation experiment environment for innovative talents of smart manufacturing between university and advanced technology companies and enterprises.This paper proposes the systematic method to establish a joint laboratory with advanced technology companies and enterprises through various forms of cooperation to get the free advanced software and data resources. The mode has been testified and applied successfully in building process and lessons of electronic equipment virtual simulation center. By this method has been used to the teaching reform project in Shaanxi Province, we have reached a cooperation with Siemens Industry software company and the user of Siemens. The innovation of this paper is to summary and practice the cooperation mode from the whole process point of view with technology companies and enterprises. The best industry-university-research model can be refereed and used in establishing industrial virtual simulation laboratory. Keywords—industry-university-research model;industrial virtual simulation laboratory;cooperation; joint laboratory; smart manufacturing

Output prediction is one of the difficult issues in production management. To overcome this difficulty, a dynamic-improved multiple linear regression model based on parameter evaluation using discrete Hopfield neural networks (DHNN) is presented. First, a traditional multiple linear regression model is established; this model takes the factors in production lifecycle (not only one phase of the production) into account, such as manufacturing resources, manufacturing process, and product rejection rate, so it makes the output prediction be more accurate. Then a static-improved model is built using the backstepping method. Finally, we obtain the dynamic-improved model based on parameter evaluation using DHNN. These three models are applied to an aviation manufacturing enterprise based on the actual data, and the results of the output prediction show that the models have practical value.