Constructing manifolds via generalized connected sums or fiber sums We can join two $S^d$'s via a connected sum to obtain $S^d\#S^d=S^d$. We can also deform a single $S^d$ via a self-connected sum (ie add a handle) to obtain $\#_\text{self}S^d=S^{d-1}\times S^1$. 
When $d=2$, the self-connected sums can generate all 2-dimensional closed oriented manifolds from $S^2$. Do we have a similar result for higher dimensions? ie can all d-dimensional closed orientable manifolds be constructed via connected sums and self-connected sums from many $S^d$'s?
If the answer is no, then we may replace connected sum by fiber sum: can all d-dimensional closed orientable manifolds be constructed via fiber sums and self-fiber sums from many $S^d$'s?
Although I asked the question in terms of manifold, here I do not require the manifold to be smooth. In other words,we may replace the term "manifold" in the above by CW complex. 
 A: Obtaining a manifold from $S^d$ by iterative connected sum is not always possible. The homology groups of the $d$-manifold $M$ change in a controlled way when taking such a connected sum with itself to obtain a manifold $M'$: we have
$$
H_k(M') = \begin{cases}
H_k(M) &\text{if }k \neq 1,d-1\\
H_k(M) \oplus \Bbb Z &\text{if }k = 1, d-1.
\end{cases}
$$
In particular, for $d > 2$ we can never obtain manifolds with torsion in their homology (like $\Bbb{RP}^3$) or with homology in middle dimensions (like $S^2 \times S^2$) by such a procedure.
A different procedure which you may be interested in, generalizing this, is called surgery theory. Ordinary connected sums involve starting with an embedding $S^0 \times D^d \to M$, removing the interior to leave a boundary $S^0 \times S^{d-1}$, and then gluing back in $D^1 \times S^{d-1}$ along this boundary to obtain a new manifold. The surgery operations work in the same way, but now starting with an embedding $S^k \times D^{d-k} \to M$, drilling out the interior to leave a boundary $S^k \times S^{d-k-1}$, and gluing in $D^{k+1} \times S^{d-k-1}$ along this boundary. Doing such a surgery along a "band" $S^1 \times D^1$ in a 2-manifold can be an instructive example.
Any closed, smooth, oriented manifold can be obtained by a sequence of surgeries, and this has been a useful tool in geometric topology. The story is more complicated for topological manifolds.
