Turbine rotors are among the most critical and highly stressed components in steam power plants. Although relatively few instances of catastrophic rotor bursts have occurred, they have resulted in lengthy forced outages and severe economic penalties to the affected utilities. To forestall the possibility of a catastrophic burst, utilities will retire the rotors affected, generally to the original equipment manufacturer's (OEM's) recommendations. The criteria and methodology for determining which rotors should be retired are proprietary and vary among manufacturers. If utilities could extend the life of these rotors by 10-20 years, then substantial savings would be made. The principal method for extending their life is weld repair and over the last 20 years, there has been a substantial increase in the number of repairs combined with the complexity of repair adopted. This project was designed to evaluate the relationship between microstructure and properties of welds on ex-service steam turbine rotor steels, after applying appropriate welding parameters and weld bead deposition sequence to minimise heat input and produce acceptable microstructures. From the project work, it is clear that the LP, IP and HP ex-service rotor forgings selected were representative of UK manufactured rotor forgings. Although never intended for weld repair, NP have demonstrated their ability to repair these steels successfully and produce weldments with acceptable microstructures and mechanical properties.

Spot welds are used to join sheet steel in automobile bodies. To ensure vehicle integrity, these welds, must be tested. Ideally, non-destructive testing would be employed. However, spot weld quality in the automotive industry is currently assessed using destructive methods. Spot welds can be tested non-destructively with ultrasound. Operators place a single crystal ultrasonic probe on a weld and interpret the returning signal to estimate the quality of the weld. However, this ultrasonic method has not been widely accepted in the past, possibly due to difficulties in manually quantifying the information contained in the signals. In an attempt to make ultrasonic testing viable for automotive use, a system has been created which automatically interprets the ultrasonic signals and classifies welds as good or bad. There are two main aspects to the systems. Firstly, echoes occurring within the signal are identified by an algorithm. This was developed after discovering the sequence in which the critical intermediate echoes occur. The second aspect of the system is classification of the spot weld based upon certain features of the identified echoes. The strength of the intermediate echoes was found to be primary source of information on weld size. Extensive experimental studies were designed and conducted to identify other potential information sources. Notably, the attenuation rate of the back-wall echoes in the signal was investigated. Most published papers in the field report that signal attenuation may be used to estimate weld size. It has been generally believed that the grain structure of the welded steel significantly increases ultrasound scattering, leading to higher attenuation is not caused by weld grain structure. The evidence gathered strongly favours weld surface irregularities as the primary cause of ultrasound attenuation in spot welds.

Three finite element model updating approaches are considered in this thesis. In the first updating approach, a line search method is used in conjunction with the existing two level Gauss-Newton approach. This two level updating approach was used to tackle the problem of shape function discretization and enabled a coarse finite element model to be updated without discretization bias. The purpose of the line search method is to place the updated estimates, which are obtained at each iteration step of the Newton method, in a superior position for the next Newton iteration. Accordingly, convergence can be achieved with the use of line searching in some non-convergent problems. The effectiveness of this updating technique is illustrated by both simulated and experimental case studies. The second approach is concerned with reduction methods for use in finite element model updating. Particular attention is paid to the performance of dynamic condensation and modal truncation schemes. In both cases, subspace iteration and the efficient computation approaches which incorporate the skyline storage method are used in conjunction with the updating method. Moreover, an efficient method is proposed for the computation of eigenvector sensitivities in principal co-ordinates. Two simulated updating problems containing over one thousand degrees of freedom are examined for a variety of cases using different forms of sensitivity data. Another area of interest in the present thesis is the modelling and updating of adhesive, welded and bolted joints. In the case of the adhesive joint, two modelling approaches, referred to as the simplified 'element strip' model and the complex 'full joint' model are examined. For the updating of the welded joint, careful parameterization is found to be critical. The use of nodal offset dimensions is shown to result in an updated joint with physical meaning. When applied to the non-linear adhesive and bolted joints, model updating will produce equivalent linearized representations. The methods and their effectiveness for mechanical joint model updating are demonstrated by experimental case studies.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.