This paper addresses the edge coloring by total labeling graph problem. This is a labeling of the vertices and edges of a graph such that the weights (colors) of the edges, defined by the sum of its label and the labels of its two endpoints, determine a proper edge coloring of the graph. We propose two integer programming formulations and derive valid inequalities which are added as cutting planes on a Branch-and-Cut framework. In order to improve the efficiency of the algorithm, we also develop initial and primal heuristics. The algorithm is tested on random instances and the computational results show that it is very effective in comparison with CPLEX. It is displayed that it reduces both the CPU time (for solved instances) and the final ...