Showing papers by "Zhigang Tian published in 2022"

A Novel Wind Speed Interval Prediction System Based on Neural Network and Multi-objective Grasshopper Optimization

Abstract: As a clean energy source, the role of wind power in the energy mix is becoming increasingly important. Reliable and high-quality wind speed prediction results are key to wind energy utilization. The traditional point prediction method cannot effectively analyze the uncertainty of wind speed, and the interval prediction model can provide the possible variation range of wind speed under a certain confidence probability and supply more uncertain information to decision makers. However, the previous interval prediction models generally ignore the random characteristics of capturing wind speed and the importance of objective selection of prediction submodels, leading to poor prediction results. To address these problems, a combined model based on data preprocessing, multi-neural network models, multi-objective optimization, and an improved interval prediction method is proposed. The model is applied to five wind speed forecasting examples in Dalian to test the prediction accuracy, multi-step prediction ability, and universality and generalization ability of the model. The experimental results show that the model proposed in this study is satisfactory for various performance evaluation indexes, has high stability and accuracy, and all the solutions obtained by the model are Pareto optimal solutions. Thus, it provides a reliable reference for the effective utilization of wind energy.

An Integrated Reliability Method with a Newly Developed Interaction Rule for Steel Pipelines with Multiple Corrosion Defects

Abstract: Although essential contributions have been made to the reliability analysis of corroded pipelines, the interacting effect between adjacent corrosion defects is rarely considered, let alone the effects of the corrosion depth and steel grade on the interacting effect. This paper proposes a new reliability method to fill the gap. First, the finite-element method and regression analysis were applied to investigate how the corrosion depth and steel grade impact the interacting effect and develop new interaction rules. Second, the new interaction rule, burst pressure model, Monte Carlo simulation (MCS), sensitivity analysis, feature scaling, and artificial neural network (ANN) were integrated to predict reliability. The proposed method combines several approaches to achieve a more accurate and efficient reliability estimation of pipelines with multiple corrosion defects than conventional assessment methods. An example is given to demonstrate the method. Results show that the limit spacing distance grows as the corrosion depth increases. The growth of the limit spacing distance of the X80 pipeline is more significant than that of the X65 pipeline. Existing interaction rules introduce conservatism to the prediction of the limit spacing distance. Two new interaction rules were developed and can realize better prediction accuracy by considering the corrosion depth and steel grade. Besides, the interacting effect significantly affects the maintenance time. The maintenance time lag between the X65 pipeline ignoring and considering the interacting effect is about 7.5 years. Different interaction rules result in different reliability descending paths. Because the new interaction rule was developed for this case, it could provide a more accurate reliability analysis. The trained ANN shows excellent prediction accuracy and high computing efficiency. The mean squared error in the reliability predicted by the ANN is 2.4×10−6. The elapsed time of the ANN prediction is about 50 times shorter than that of the MCS.Practical ApplicationsIn engineering practice, corrosion defects usually occur in patches, and the interacting effect between adjacent defects impairs the structural reliability of corroded pipelines. The corrosion depth and steel grade are found to significantly affect the interacting effect and limit spacing distance. The new interaction rules can be applied to achieve a more accurate limit spacing distance estimation by considering the effects of the corrosion depth and steel grade. The interacting effect plays a vital role in the reliability analysis of corroded pipelines. The integrated reliability method proposed in this paper considers the interacting effect and interaction rule. In practical application, once the pipeline and defect information, such as wall thickness and corrosion depth, is gathered, the proposed method can be used to realize a more accurate and effective reliability analysis of pipelines with multiple corrosion defects. Thus, the proposed reliability method and results of this paper are beneficial for the integrity management of corroded pipelines.

Speed Adaptive Gates for Deep-Learning-based Fault Classification of Rotating Machinery under Varying Speed Conditions: Comparison of Implementations

Abstract: Deep-learning-based rotating machinery fault classification often suffers from the problem of speed induced fault information imbalance when applied to varying speed conditions. The speed adaptive gate (SAG) is an effective auxiliary branch that assists existing deep learning models addressing that problem. This paper presents four types of implementations for SAGs, i.e., the element-wise, channel-wise, time-wise and sample-wise, and compares their fault classification performances with a baseline Convolutional Neural Network (CNN) over two experimental datasets including a planetary gearbox dataset and a fixed-shaft gearbox dataset. Results show that the element-wise achieves the highest fault classification accuracy while the time-wise returns the least. But element-wise consumes about 20% more training CPU time than that of the baseline CNN. A good trade-off is the sample-wise with average speed as the input to the SAG, which obtains fair accuracy improvement, but spends only about 5% more training time than the baseline CNN.