Possible bug: significant difference in integration point stresses and stress at element's centroid

[coolFL]

Hello all, 

I did one elastic-plastic simulation using Abaqus/FEA. I used C3D20T (hexahedral quad) solid element which has 27 integration points and 20 nodes. For one of the element, the stress that I got at the centroid is completely different from the stresses that I got at the integration points. Can somebody please tell me how Abaqus determines stresses at the centroid of the element from its integration pt. values? For reference, stresses at the integration points and centroid of this element are as follows:

 It. pt. 1 :  2929.26 MPa
It. pt. 2  :  2899.97 MPa
It. pt. 3 : 2744.67 MPa
It. pt. 4 : 2959.99 MPa
It. pt. 5 : 2923.19 MPa
It. pt. 6 : 2892.74 MPa
It. pt. 7 : 3008.94 MPa
It. pt. 8: 2946.87 MPa
It. pt. 9 : 2914.68 MPa
It. pt. 10 : 2901 MPa
It. pt. 11 : 2622.9 MPa
It. pt. 12 : 2328.21 MPa
It. pt. 13 : 2928.13 MPa
It. pt. 14 : 2895.74 MPa
It. pt. 15 : 2553.81 MPa
It. pt. 16 : 2956.9 MPa
It. pt. 17 : 2919.5 MPa
It. pt. 18 : 2802.65 MPa
It. pt. 19 : 2593.25 MPa
It. pt. 20 : 2239.77 MPa
It. pt. 21 : 1997.29 MPa
It. pt. 22 : 2913.44 MPa
It. pt. 23 : 2534.04 MPa
It. pt. 24 : 2190.04 MPa
It. pt. 25 : 2947.21 MPa
It. pt. 26 : 2866.59 MPa
It. pt. 27 : 2427.03 MPa

At the centroid, stress is 3402.07 MPa

Thank you in advance, 

Nik