A wavelet approach for mathematical physics is considered in the paper. Application of wavelet basises for solving problems of mathematical physics - a relatively new promising area of science. We investigate a approximation of function using WAVE, MHAT, DOG, MORLET and Daubechies wavelets. A effect of influence of scale of detalization and type of mother wavelet to the accuracy of the approximation studied in detail. For wavelets WAVE, MHAT, DOG, MORLET approximation quality depends from the choice a one of them: for the example function the highest rate of convergence is marked for wavelet WAVE and MORLET, worse results for DOG and MHAT. It should also be noted that after a certain level increase of accuracy does not occur. This is due, apparently, the implementation details of the algorithm, accurate computation of integrals. Orthogonal wavelets (HAAR (Dl), D2, DIO) show a more stable behavior: with increasing order of Daubechies wavelet convergence rate increases. The method for boundary eigenvalue promlem is proposed. Depending of number of layers the method calculate some number of eigenvalues. Thus, the method allows to filter the eigenvalues corresponding to eigenfunctions of given scale.