Fault Diagnosis in Gas Turbine Based on Neural Networks: Vibrations Speed Application
The diagnosis of faults and failures in industrial systems is becoming increasingly essential. This work proposes the development of a fault diagnostics system based on artificial intelligence technique, using neural networks applied to a GE MS3002 gas tu
- PDF / 306,817 Bytes
- 11 Pages / 439.37 x 666.142 pts Page_size
- 75 Downloads / 220 Views
Abstract The diagnosis of faults and failures in industrial systems is becoming increasingly essential. This work proposes the development of a fault diagnostics system based on artificial intelligence technique, using neural networks applied to a GE MS3002 gas turbine. This technique with its generalization and memory skills provides an effective diagnostic tool for the examined system.
Keywords Diagnosis of defects Neural networks of residues Gas turbine Vibration
Modeling
Generation
1 Introduction The dream of creating an intelligence system is presented for a long time in the human imagination. Several applications in this direction have been conducted by scientists with the use of neural networks as a decision support tool (Hafaifa et al. 2011a, b, c, 2013, 2014b, c, 2015a, b; Chen and Lee 2002; Tsai and Chang 1995; Halimi et al. 2014a, b; Eshati et al. 2013; Temurtas 2009; Galindo et al. 2013; McGhee et al. 1997; Kim et al. 2011; Leger et al. 1998; Owen 2012; Nikpey et al. 2014). The creation of neural systems requires a thorough knowledge of several technical aspects (Sanaye and Tahani 2010; Wahba and Nawar 2013; Wu et al. 2011; Yang et al. 2000; Kim et al. 2010; Zhang et al. 1994). In the diagnosis of M. Ben Rahmoune A. Hafaifa (&) Applied Automation and Industrial Diagnostics Laboratory, University of Djelfa, 17000 Djelfa, DZ, Algeria e-mail: [email protected] M. Ben Rahmoune e-mail: [email protected] M. Guemana Faculty of Science and Technology, University of Médéa, 26000 Médéa, DZ, Algeria e-mail: [email protected] © Springer International Publishing Switzerland 2017 T. Fakhfakh et al. (eds.), Advances in Acoustics and Vibration, Applied Condition Monitoring 5, DOI 10.1007/978-3-319-41459-1_1
1
2
M. Ben Rahmoune et al.
faults and failures of industrial systems, this area is becoming more and more complex. Artificial neural networks, are now a well understood and managed data processing technique that allows the engineer to extract, in many situations, the most relevant information of the data it has: control processes, prediction properties, modeling, pattern recognition. It is from the assumption that intelligent behavior emerges from the structure and behavior of brain basics, the artificial neural networks have developed. Each artificial neuron is a functional unit. He receives a variable number of inputs from upstream neuron. Each one of these entries is associated with a weight ω representative of the strength of the connection. Each functional unit has a single output, which branches to supply follows, a variable number of downstream neuron (Eshati et al. 2013). In this context there are several applications of neural networks for fault diagnosis and in particular, for the diagnosis of faults of several industrial applications. In this work, vibration measurements parameters are used for the development of a vibration monitoring system based on neural networks system. These parameters can be the displacement, velocity and acceleration; in time or frequenc
Data Loading...
