The determination of individual stress values in the multiaxial residual stress distributions found in toughened glass is a problem of major practical importance in the glass industry. In spite of many years of development, conventional experimental stress analysis techniques have not been able to provide a generally applicable means of obtaining this information. However, a unique method has now been developed for this purpose, based on the principles of magnetophotoelasticity, in which the combined optical retardation effects due to (i) the stress-birefringence of the material and (ii) a superimposed magnetic field, are utilized.The paper summarizes the basic theory of magnetophotoelasticity and describes experimental verification work using simple uniaxial systems of constant and linearly varying stress. Two important developments follow. First, a solution technique is developed whereby principal stress differences can be determined at any position through the thickness of a toughened glass plate, for an assumed biaxial residual stress distribution consisting of a combination of parabolic and linear stress variations. Second, an oblique incidence technique is employed to obtain individual stresses from the previously determined principal stress differences.A polariscope, designed and built for the implementation of the technique at any position over the area of an automotive windscreen, is described. Typical experimental results are presented and shown to compare well with the limited information available from conventional techniques.

In an unstressed orthorhombic material the shear wave velocity is invariant under an interchange of the propagation and polarisation directions, if these directions lie along the principal texture axes of the material. In the presence of a stress with principal stress directions coincident with the principal texture axes, this symmetry is broken and the change in the velocity when the propagation and polarisation directions are interchanged is given by the difference in stresses along the propagation and polarisation directions. In this paper, the effect of a misalignment of the symmetry and sample planes on this relationship has been examined. For austenitic welds, an error in the determination of stress of the order of yield can occur for tilts of 5 degrees . However, for structural steels with textures very much less than the theoretical maximum, a determination of the stress present to an accuracy of 10% of yield remains possible using this technique.