Application of least square method to arbitrary-order problems with separated boundary conditions (Q955072)

The author presents a B-spline method to approximate by least-squares solutions of higher ordinary differential equations (i) \(g(t,y(t),y'(t),\dots, y^{(n)}(t))= 0\), \(a\leq t\leq b\), with separated boundary conditions (ii) \(\sum^{m-1}_{j=1} \alpha_{ij} y^{(j)}(a_i)= A_0\), \(i= 1,\dots, m\), \(a\leq a_1\leq a_2\leq\cdots\), \(a_m\leq b\). The method generates a sequence of functions \[ v_k(t)= \sum_{i\in S(k)} c_i B_{ki}(t),\;S(k)= (-m,\dots, (m+1)2^{m-1}- 1)), \] \(k\in\mathbb{N}\), where \(B_{ki}\) are specially chosen \(m\)th degree B-splines. Given \(\varepsilon> 0\), choose \(k\in\mathbb{N}\) and \(c_i(k)\) so that \(v_k(t)\) satisfies (ii) and \[ \int^b_a g(t,v_k(t),\dots, v^{(m)}_k(t))^2\,dt<\varepsilon. \] The sequence \(v_k(t)\) converges to a solution of (i), (ii), \(k\to\infty\). Results of computations for eight test problems are given which include linear equations of order \(m= 5,8,9,12\) and nonlinear ones of order \(m= 2,4\).