I want to code in matlab to calculate a cubic thermal expansion problem(length of the cube is 10 ,temperature change is 100, and CTE is 0.0001). I code to calculate the mass matrix, stiffness matrix and thermal load, these three matrixes or vector are the same as abaqus(I have already considered the hourglass problem, i use the deflaut setting), but when i want to calculate the node displacement, i use KU =F,then U=K^-1*F to calculate node displacement by matlab code, it is wrong, the abaqus result and real result are the same 0.1 0.1 0.1 at the node(Obvious this result), but my result is 0.0984,0.0966,0.0987. Although my matrix and load are same as abaqus, how abaqus calculate KU =F in linear problem.Thanks. It is not the problem of inv(K) orK\\ or K^-1，they have the same results as mine.