Experimental results of stress measurements carried out on the front and back faces of welded plates are shown in figure 3 and figure 4. Stress distribution curves both for longitudinal and transverse residual stresses are presented. The stress state on the initial surface is characterised by compressive residual stress at the weld seam and tensile stresses in the HAZ and base metal.
Since the residual stress are selfequilibrium, results presented in the figures 3 and 4 have to satisfy to equilibrium equation:
 (7)

where A is the area of crosssection perpendicular to analysed stress component. For longitudinal e transverse residual stresses the equation (7) is transformed to following two equations:
 (8)

 (9)

To satisfy the equilibrium of residual stresses it is necessary to know the stress variation along Zcoordinate (see figure 1). If one assumes that the stresses within weld seam, HAZ and base metal are the same as on the outer surface then the stress distribution curve for the longitudinal direction is satisfied for the equilibrium equation because of variation of stress sign along Xcoordinate. For transverse residual stress (for example at the centre of the weld seam) the equilibrium equation is not satisfied because st can not change in sign along the ycoordinate. In practice, stress measurement at point 11, near the edge of weld seam, show that the value of transverse stress at this point is equal to
s= 150Mpa and it is not very different from stress value at the centre of seam.
Fig 3: Residual stress distribution on front face of welded plate:
1longitudinal stresses; 2transverse stresses.

Therefore, satisfaction of the equilibrium equation for transverse residual stress requires the modification of the compressive stress acting at the outer surface of the weld seam to a tensile residual stress within interior regions of the seam.. This conclusion means that stress measurements on the surface of a welded joint is not sufficient to examine residual stresses arising after welding. To verify this point of view it is necessary to determine the stress distribution throughout the depth of the analysed plate. This can be accomplished by means of removing of surface layer and carrying out stress measurments on a new surface. It is clear that in the case of layer removal by machining or grinding it is necessary to undertake electropolishing to remove additional surface layer distortion by machining.
Stress measurements after surface removal are presented in figure 5. Curve 1 of this figure shows that compressive residual stresses in the transverse direction acting at the centre of weld seam become tensile stresses that reach s_{t} = 250 MPa at 2,5 mm from the outer surface. The curve 2 in figure 5 shows the inhomogeneous stress distribution along the weld seam structure. At the centre of the weld pass (x = 0 mm) the stress value is higher than at the border of weld pass (x = 2 mm).
Fig 4: Residual stress distribution on back face of welded plate:
1longitudinal stresses; 2transverse stresses.

Fig 5: Stress distributions along Z coordinate (curve 1) and X coordinate (curve 2).

Experimental results presented in figures35 do not contradict the theory of the origin of residual stresses after welding. Accordance to [5] the principal sources of residual stresses after welding are:
 shrinkage;
 quenching;
 phase transformation.
Shrinkage provokes the appearance of tensile stresses, whereas quenching and phase transformation cause compressive stresses at the weld seam.
Initial compressive residual stresses in the weld seam indicate that their main sources are quenching or phase transformation. The stress distribution curve presented in figure 5 indicates that quenching is the predominant cause of the creation of a residual stress state. In practice, the stress state of quenched part is characterised by compressive residual stresses on the outer surface and tensile stresses in the interior region.