Hui Zhai, Wei Xiong, Fujin Li, Jie Yang, Dongyan Su and Yongjun Zhang
The prediction of by-product gas is an important guarantee for the full utilization of resources. The purpose of this research is to predict gas consumption to provide a basis for…
Abstract
Purpose
The prediction of by-product gas is an important guarantee for the full utilization of resources. The purpose of this research is to predict gas consumption to provide a basis for gas dispatch and reduce the production cost of enterprises.
Design/methodology/approach
In this paper, a new method using the ensemble empirical mode decomposition (EEMD) and the back propagation neural network is proposed. Unfortunately, this method does not achieve the ideal prediction. Further, using the advantages of long short-term memory (LSTM) neural network for long-term dependence, a prediction method based on EEMD and LSTM is proposed. In this model, the gas consumption series is decomposed into several intrinsic mode functions and a residual term (r(t)) by EEMD. Second, each component is predicted by LSTM. The predicted values of all components are added together to get the final prediction result.
Findings
The results show that the root mean square error is reduced to 0.35%, the average absolute error is reduced to 1.852 and the R-squared is reached to 0.963.
Originality/value
A new gas consumption prediction method is proposed in this paper. The production data collected in the actual production process is non-linear, unstable and contains a lot of noise. But the EEMD method has the unique superiority in the analysis data aspect and may solve these questions well. The prediction of gas consumption is the result of long-term training and needs a lot of prior knowledge. Relying on LSTM can solve the problem of long-term dependence.
Details
Keywords
Qingshan Wang, Dongyan Shi, Qian Liang and Fuzhen Pang
The purpose of this work is to apply the Fourier–Ritz method to study the vibration behavior of the moderately thick functionally graded (FG) parabolic and circular panels and…
Abstract
Purpose
The purpose of this work is to apply the Fourier–Ritz method to study the vibration behavior of the moderately thick functionally graded (FG) parabolic and circular panels and shells of revolution with general boundary conditions.
Design/methodology/approach
The modified Fourier series is chosen as the basis function of the admissible functions of the structure to eliminate all the relevant discontinuities of the displacements and their derivatives at the edges, and the vibration behavior is solved by means of the Ritz method. The complete shells of revolution can be achieved by using the coupling spring technique to imitate the kinematic compatibility and physical compatibility conditions of FG parabolic and circular panels at the common meridian of θ = 0 and 2π. The convergence and accuracy of the present method are verified by other contributors.
Findings
Some new results of FG panels and shells with elastic restraints, as well as different geometric and material parameters, are presented and the effects of the elastic restraint parameters, power-law exponent, circumference angle and power-law distributions on the free vibration characteristic of the panels are also presented, which can be served as benchmark data for the designers and engineers to avoid the unpleasant, inefficient and structurally damaging resonant.
Originality/value
The paper could provide the reference for the research about the moderately thick FG parabolic and circular panels and shells of revolution with general boundary conditions. In addition, the change of the boundary conditions can be easily achieved by just varying the stiffness of the boundary restraining springs along all the edges of panels without making any changes in the solution procedure.