A compensation auto-associative neural network for overcoming smearing effects

被引：0
|
作者：
Ren S. [1 ,2 ]
Xiao J. [1 ,2 ]
Si F. [1 ,2 ]
Cao Y. [1 ,2 ]
Chen J. [1 ,2 ]
机构：
[1] Key Laboratory of Energy Thermal Conversion and Control of Ministry of Education, Southeast University, Nanjing
[2] School of Energy and Environment, Southeast University, Nanjing
来源：
Dongnan Daxue Xuebao (Ziran Kexue Ban)/Journal of Southeast University (Natural Science Edition) | 2020年 / 50卷 / 04期
关键词：
Auto-associative neural network; Fault diagnosis; Residual compensation; Smearing effects;
D O I：
10.3969/j.issn.1001-0505.2020.04.016
中图分类号：
学科分类号：
摘要：
To overcome the smearing effect problem of the basic auto-associative neural network(AANN) that may reduce the reconstruction accuracy, a new compensation auto-associative neural network (CAANN) is proposed. In CAANN, a new compensation layer is introduced into the testing process of the basic AANN in order to develop an adjustment mechanism between the input layer and the residual space, and the residual compensation magnitude is calculated by the gradient descent algorithm. Then, a strategy of residual compensation for both single and multiple variable(s) is implemented to calculate the optimized compensation magnitude and direction by comparing the squared prediction error(SPE) after compensation, and further the source of anomalies is pinpointed and the reconstructed values are obtained. The effectiveness of the proposed method is evaluated on a validation example and an industrial example. The results demonstrate that the proposed CAANN can effectively inhibit the drawbacks of "smearing effects" in case of both large amplitude anomalies and multipoint concurrent anomalies without prior knowledge, showing a better performance than AANN and principal component analysis(PCA) methods for fault diagnosis and data reconstruction. © 2020, Editorial Department of Journal of Southeast University. All right reserved.
引用
收藏
页码：712 / 720
页数：8
相关论文
共 21 条
  • [1] Heo S, Lee J H., Parallel neural networks for improved nonlinear principal component analysis, Computers & Chemical Engineering, 127, pp. 1-10, (2019)
  • [2] Ghosh A, Wang G N, Lee J., A novel automata and neural network based fault diagnosis system for PLC controlled manufacturing systems, Computers & Industrial Engineering, 139, (2020)
  • [3] Hines J W, Uhrig R E, Wrest D J., Use of autoassociative neural networks for signal validation, Journal of Intelligent and Robotic Systems, 21, pp. 143-154, (1998)
  • [4] Kramer M A., Autoassociative neural networks, Computers & Chemical Engineering, 16, 4, pp. 313-328, (1992)
  • [5] Kramer M A., Nonlinear principal component analysis using autoassociative neural networks, AIChE Journal, 37, 2, pp. 233-243, (1991)
  • [6] Gautam C, Ravi V., Counter propagation auto-associative neural network based data imputation, Information Sciences, 325, pp. 288-299, (2015)
  • [7] Zhou JM, Zhang CC, Zhang L, MachineDesign&Research, 35, 1, pp. 96-99, (2019)
  • [8] Xiao H J, Huang D P, Pan Y P, Et al., Fault diagnosis and prognosis of wastewater processes with incomplete data by the auto-associative neural networks and ARMA model, Chemometrics and Intelligent Laboratory Systems, 161, pp. 96-107, (2017)
  • [9] Elnour M, Meskin N, Al-Naemi M., Sensor data validation and fault diagnosis using auto-associative neural network for HVAC systems, Journal of Building Engineering, 27, (2020)
  • [10] Shang C, Ji H Q, Huang X L, Et al., Generalized grouped contributions for hierarchical fault diagnosis with group Lasso, Control Engineering Practice, 93, (2019)
123