Nalinaksh S. Vyas

What Is Engineering Asset Management

Definitions of asset management tend to be broad in scope, covering a wide variety of areas including general management, operations and production arenas and, financial and human capital aspects. While the broader conceptualisation allows a multifaceted investigation of physical assets, the arenas constitute a multiplicity of spheres of activity. We define engineering asset management in this paper as the total management of physical, as opposed to financial, assets. However, engineering assets have a financial dimension that reflects their economic value and the management of this value is an important part of overall engineering asset management. We also define more specifically what we mean by an “engineering asset” and what the management of such an asset entails. Our approach takes as its starting point the conceptualisation of asset management that posits it as an interdisciplinary field of endeavour and we include notions from commerce and business as well as engineering. The framework is also broad, emphasising the life-cycle of the asset. The paper provides a basis for analysing the general problem of physical asset management, relating engineering capability to economic cost and value in a highly integrated way.

Artificial neural network design for fault identification in a rotor-bearing system

Abstract A neural network simulator built for prediction of faults in rotating machinery is discussed. A back-propagation learning algorithm and a multi-layer network have been employed. The layers are constituted of nonlinear neurons and an input vector normalization scheme has been built into the simulator. Experiments are conducted on an existing laboratory rotor–rig to generate training and test data. Five different primary faults and their combinations are introduced in the experimental set-up. Statistical moments of the vibration signals of the rotor-bearing system are employed to train the network. Network training is carried out for a variety of inputs. The adaptability of different architectures is investigated. The networks are validated for test data with unknown faults. An overall success rate up to 90% is observed.

Non-linear parameter estimation in multi-degree-of-freedom systems using multi-input Volterra series

Abstract Functional form input–output representation through Volterra series has been widely used for non-linear system analysis and non-parametric system identification. Recent research work shows that the series representation can be suitably employed for parametric identification also. However, the classical Volterra series is based on a single-input and its application is limited to analysis and identification of single-degree-of-freedom system only. The concept of single-input Volterra series has been extended to multi-input Volterra series by Worden et al. through definition of direct and cross-kernels. The present study employs the multi-input Volterra series and develops a structured response representation of various harmonics under multi-input harmonic excitations. Kernel synthesis formulations are developed for a polynomial form non-linearity with general square and cubic terms. It is shown that higher-order direct and cross-kernel transforms are functions of the first-order kernel transforms and the non-linear parameter vectors. A parameter estimation procedure based on recursive iteration is suggested and illustrated for a two-degree-of-freedom system with square and cubic stiffness non-linearity. Numerical simulations and error analysis are presented for typical rotor-bearing system parameters.

Non-linear parameter estimation with Volterra series using the method of recursive iteration through harmonic probing

Volterra series provides a platform for non-linear response representation and definition of higher order frequency response functions (FRFs). It has been extensively used in non-parametric system identification through measurement of first and higher order FRFs. A parametric system identification approach has been adopted in the present study. The series response structure is explored for parameter estimation of polynomial form non-linearity. First and higher order frequency response functions are extracted from the measured response harmonic amplitudes through recursive iteration. Relationships between higher order FRFs and first order FRF are then employed to estimate the non-linear parameters. Excitation levels are selected for minimum series approximation error and the number of terms in the series is controlled according to convergence requirement. The problem of low signal strength of higher harmonics is investigated and a measurability criterion is proposed for selection of excitation level and range of excitation frequency. The procedure is illustrated through numerical simulation for a Duffing oscillator. Robustness of the estimation procedure in the presence of measurement noise is also investigated.

EQUATIONS OF MOTION OF A BLADE ROTATING WITH VARIABLE ANGULAR VELOCITY

Abstract The equations of motion of a blade mounted on a disk rotating with variable angular velocity are derived. The acceleration of the disk is taken as constant and the calculus of variations is employed to obtain the equation governing the ensuing free vibrations. Coriolis forces are included in the derivation and the higher order terms due to shear deflection and rotary inertia are also considered.

Application of Volterra and Wiener Theories for Nonlinear Parameter Estimation in a Rotor-Bearing System

Volterra and Wiener theories provide the concepts of linear,bilinear, tri-linear, etc., kernels, which upon convolution with theexcitation force, can be employed to represent the response of anonlinear system. Based on these theories, higher-order frequencyresponse functions (FRFs) are employed to estimate the nonlinearstiffness of rolling element bearings, supporting a rigid rotor. Therotor-bearing assembly is idealized as a single-degree-freedom system,with cubic nonlinearity. The analysis involves a third-order kernelrepresentation of the system response. The first and third-order kerneltransforms are extracted from the measurements of the appliedwhite-noise excitation and the resultant response. A third-order kernelfactor is synthesized from this first-order kernel and is processedalong with the third-order kernel for estimation of the nonlinearparameter. Damping is assumed to be linear in the analysis. Theprocedure is demonstrated through measurements on a laboratory test rig.

Determination of Blade Stresses Under Constant Speed and Transient Conditions With Nonlinear Damping

Determination of resonant stresses is an important step in the life estimation of turbomachine blades. Resonance may occur either at a steady operating speed or under transient conditions of operation when the blade passes through a critical speed. Damping plays a significant role in limiting the amplitudes of vibration and stress values. The blade damping mechanism is very complex in nature, because of interfacial slip, material hysteresis, and gas dynamic damping occurring simultaneously. In this paper, a numerical technique to compute the stress response of a turbine blade with nonlinear damping characteristics, during steady and transient operations of the rotor, is presented. Damping is defined as a function of vibratory mode, rotor speed, and strain amplitude. The technique is illustrated by computing the stress levels at resonant rotor speeds for typical operation of a turbomachine.

Nonlinear bearing stiffness parameter estimation in flexible rotor–bearing systems using Volterra and Wiener approach

Higher order frequency response functions based on Volterra and Wiener series are explored for the inverse problem of stiffness estimation of a flexible rotor supported in nonlinear bearings. The Volterra series has been employed by researchers earlier for the identification of higher order kernels of nonlinear systems through a non-parametric approach. The present study investigates the possibility of employing these kernels for parameter estimation of the system. Numerical simulation has been carried out for a system with three degrees of freedom and with cubic nonlinearity in stiffness. A frequency domain has been adopted for the identification of higher order kernels. The procedure involves extraction of Wiener kernels from the response of the system to a Gaussian white noise excitation. Volterra kernels are in turn synthesised from the Wiener kernels. In addition to direct kernels, the system under consideration, requires definitions of cross-kernels and their estimation. Expressions for the cross and direct kernels are constructed in the frequency domain. A set of third-order kernel factors are algebraically and graphically synthesised from the measured first-order kernels. These third-order kernel factors are then processed with the measured third-order kernels for nonlinear parameter estimation. Damping is taken to be linear in the analysis. The procedure is illustrated through numerical simulation. The assumptions involved and the approximations are discussed.

Nonlinear Parameter Estimation in Rotor-Bearing System Using Volterra Series and Method of Harmonic Probing

Volterra series provides a structured analytical platform for modeling and identification of nonlinear systems. The series has been widely used in nonparametric identification through higher order frequency response functions or FRFs. A parametric identification procedure based on recursive evaluation of response harmonic amplitude series is presented here. The procedure is experimentally investigated for a rotor-bearing system supported in rolling element bearings. The estimates of nonlinear bearing stiffness obtained from experimentation have been compared with analytical values and experimental results of previous works.

Dynamic stress analysis and a fracture mechanics approach to life prediction of turbine blades

Emerging blade technologies are finding it increasingly essential to correlate blade vibrations to blade fatigue in order to assess the residual life of existing blading and for development of newer designs. In this paper an analytical code for dynamic stress analysis and fatigue life prediction of blades is presented. The life prediction algorithm is based on a combination method, which combines the local strain approach to predict the initiation life and fracture mechanics approach to predict the propagation life, to estimate the total fatigue life. The conventional stress based approach involving von Mises theory along with S-N-Mean stress diagram suffers from the drawback that it does not make allowance for the possibility of development of plastic strain zones, especially in cases of low cycle fatigue. In the present paper, strain life concepts are employed to analyse the crack initiation phenomenon. Dynamic and static stresses incurred by the blade form inputs to the life estimation algorithm. The modeling is done for a general tapered, twisted and asymmetric cross section blade mounted on a rotating disc at a stagger angle. Blade damping is non-linear in nature and a numerical technique is employed for estimation of blade stresses under typical nozzle excitation. Critical cases of resonant conditions of blade operation are considered. Neubers rule is applied to the dynamic stresses to obtain the elasto-plastic strains and then the material hysteresis curve is used to iteratively solve for the plastic stress. Static stress effects are accounted for and crack initiation life is estimated by solving the strain life equation. Crack growth formulations are then applied to the initiated crack to analyse the propagation of crack leading to failure. The engineering approximations involved are stated and the algorithm is numerically demonstrated for typical conditions of blade operations.