Develop Reference

Fastest way to solve non-negative linear diophantine equations

Let A be a list of n lists of m non-negative integers, such that for all j there is i with A[i][j] nonzero. Let V be a list of m positive integers.
Question: What is the fastest way to find all the lists X of n non-negative integers such that for all i then sum_j A[i][j] X[j] = V[i]?
The assumptions implies that the number of solutions is finite. See below an example of A and V, with 5499 solutions X.
Let me reformulate the problem using matrix and vector. Let A be a n-by-m matrix with non-negative integral entries and without zero column. Let V be a vector with positive integral entries. What is the fastest way to find all the vectors X, with non-negative integral entries, such that AX=V?
The usual functions for solving such a system may underuse the non-negativity. To prove so, I wrote a brute-force code finding all the solutions of such a system and applied it to an example (see below, and these crossposts on mathoverflow and on ask.sagemath), but I'm still looking for something significantly faster than this; in fact I'm looking for the fastest way.
Example
Here is the kind of system I need to solve (where x_i is non-negative integral), but with possibly more equations and variables.
[
5*x0 + 5*x1 + 5*x2 + 6*x3 + 7*x4 + 7*x5 == 24,
5*x1 + 7*x10 + 5*x6 + 5*x7 + 6*x8 + 7*x9 == 25,
5*x11 + 6*x12 + 7*x13 + 7*x14 + 5*x2 + 5*x7 == 25,
5*x12 + 6*x15 + 7*x16 + 7*x17 + 5*x3 + 5*x8 == 30,
5*x13 + 6*x16 + 7*x18 + 7*x19 + 5*x4 + 5*x9 == 35,
5*x10 + 5*x14 + 6*x17 + 7*x19 + 7*x20 + 5*x5 == 35,
5*x1 + 7*x10 + 5*x6 + 5*x7 + 6*x8 + 7*x9 == 25,
5*x21 + 5*x22 + 6*x23 + 7*x24 + 7*x25 + 5*x6 == 24,
5*x22 + 5*x26 + 6*x27 + 7*x28 + 7*x29 + 5*x7 == 25,
5*x23 + 5*x27 + 6*x30 + 7*x31 + 7*x32 + 5*x8 == 30,
5*x24 + 5*x28 + 6*x31 + 7*x33 + 7*x34 + 5*x9 == 35,
5*x10 + 5*x25 + 5*x29 + 6*x32 + 7*x34 + 7*x35 == 35,
5*x11 + 6*x12 + 7*x13 + 7*x14 + 5*x2 + 5*x7 == 25,
5*x22 + 5*x26 + 6*x27 + 7*x28 + 7*x29 + 5*x7 == 25,
5*x11 + 5*x26 + 5*x36 + 6*x37 + 7*x38 + 7*x39 == 24,
5*x12 + 5*x27 + 5*x37 + 6*x40 + 7*x41 + 7*x42 == 30,
5*x13 + 5*x28 + 5*x38 + 6*x41 + 7*x43 + 7*x44 == 35,
5*x14 + 5*x29 + 5*x39 + 6*x42 + 7*x44 + 7*x45 == 35,
5*x12 + 6*x15 + 7*x16 + 7*x17 + 5*x3 + 5*x8 == 30,
5*x23 + 5*x27 + 6*x30 + 7*x31 + 7*x32 + 5*x8 == 30,
5*x12 + 5*x27 + 5*x37 + 6*x40 + 7*x41 + 7*x42 == 30,
5*x15 + 5*x30 + 5*x40 + 6*x46 + 7*x47 + 7*x48 == 35,
5*x16 + 5*x31 + 5*x41 + 6*x47 + 7*x49 + 7*x50 == 42,
5*x17 + 5*x32 + 5*x42 + 6*x48 + 7*x50 + 7*x51 == 42,
5*x13 + 6*x16 + 7*x18 + 7*x19 + 5*x4 + 5*x9 == 35,
5*x24 + 5*x28 + 6*x31 + 7*x33 + 7*x34 + 5*x9 == 35,
5*x13 + 5*x28 + 5*x38 + 6*x41 + 7*x43 + 7*x44 == 35,
5*x16 + 5*x31 + 5*x41 + 6*x47 + 7*x49 + 7*x50 == 42,
5*x18 + 5*x33 + 5*x43 + 6*x49 + 7*x52 + 7*x53 == 48,
5*x19 + 5*x34 + 5*x44 + 6*x50 + 7*x53 + 7*x54 == 49,
5*x10 + 5*x14 + 6*x17 + 7*x19 + 7*x20 + 5*x5 == 35,
5*x10 + 5*x25 + 5*x29 + 6*x32 + 7*x34 + 7*x35 == 35,
5*x14 + 5*x29 + 5*x39 + 6*x42 + 7*x44 + 7*x45 == 35,
5*x17 + 5*x32 + 5*x42 + 6*x48 + 7*x50 + 7*x51 == 42,
5*x19 + 5*x34 + 5*x44 + 6*x50 + 7*x53 + 7*x54 == 49,
5*x20 + 5*x35 + 5*x45 + 6*x51 + 7*x54 + 7*x55 == 48
]
Here are explicit A and V from above system (in list form):
A=[
[5,5,5,6,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,5,0,0,0,0,5,5,6,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,5,0,0,0,0,5,0,0,0,5,6,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,5,0,0,0,0,5,0,0,0,5,0,0,6,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,5,0,0,0,0,5,0,0,0,5,0,0,6,0,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,5,0,0,0,0,5,0,0,0,5,0,0,6,0,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,5,0,0,0,0,5,5,6,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,5,6,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,5,6,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,5,0,0,6,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,5,0,0,6,0,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,5,0,0,6,0,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,5,0,0,0,0,5,0,0,0,5,6,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,5,6,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,5,6,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,5,0,0,6,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,5,0,0,6,0,7,7,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,5,0,0,6,0,7,7,0,0,0,0,0,0,0,0,0,0],
[0,0,0,5,0,0,0,0,5,0,0,0,5,0,0,6,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,5,0,0,6,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,5,0,0,6,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,6,7,7,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,6,0,7,7,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,6,0,7,7,0,0,0,0],
[0,0,0,0,5,0,0,0,0,5,0,0,0,5,0,0,6,0,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,5,0,0,6,0,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,5,0,0,6,0,7,7,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,6,0,7,7,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,6,0,0,7,7,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,6,0,0,7,7,0],
[0,0,0,0,0,5,0,0,0,0,5,0,0,0,5,0,0,6,0,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,5,0,0,6,0,7,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,5,0,0,6,0,7,7,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,6,0,7,7,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,6,0,0,7,7,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,6,0,0,7,7]
]
V=[24,25,25,30,35,35,25,24,25,30,35,35,25,25,24,30,35,35,30,30,30,35,42,42,35,35,35,42,48,49,35,35,35,42,49,48]
Computation
I wrote a brute-force code finding all the solutions of such a system, then applied it to A, V above. It took 12 seconds to find all 5499 solutions. I'm looking for something significantly faster than this.
sage: %time LX=NonnegativeSolverPartition(A,V)
CPU times: user 11.8 s, sys: 0 ns, total: 11.8 s
Wall time: 11.8 s
sage: len(LX)
5499
Note that the time reduces to 3 seconds with PyPy3, but it is still too slow for other (bigger) such systems I need to solve.
Code
Here is my (python) code, improved by Peter Taylor (see his comment):
def NonnegativeSolverPartition(A,V):
WB = WeakBasis(A)
VB = VarBound(A,V)
PP = []
for i, ll in WB:
L = tuple(A[i][j] for j in ll)
B = tuple(VB[j] for j in ll)
PP.append(WeightedPartitionSolver(L, B, V[i]))
return list(NonnegativeSolverPartitionInter(A, V, PP, WB, [-1] * len(A[0])))
def NonnegativeSolverPartitionInter(A, V, PP, WB, X):
if any(len(P) > 1 for P in PP):
_, ii = min((len(P), i) for i, P in enumerate(PP) if len(P) > 1)
for p in PP[ii]:
PPP = PP[:ii] + [[p]] + PP[ii+1:]
Fi = Filter(PPP, list(X), WB)
if Fi:
PPPP, XX = Fi
yield from NonnegativeSolverPartitionInter(A, V, PPPP, WB, XX)
else:
assert -1 not in X
yield X
def WeakBasis(A):
return tuple(enumerate([j for j, tmp in enumerate(row) if tmp] for row in A))
def WeightedPartitions(ws, n):
def inner(ws, n):
if n == 0:
yield (0,) * len(ws)
elif ws:
w = ws[0]
lim = n // w
ws = ws[1:]
for i in range(lim + 1):
for tl in inner(ws, n - w * i):
yield (i,) + tl
return list(inner(ws, n))
def VarBound(A,V):
nvars = len(A[0])
# Solve the individual constraints and then intersect the solutions.
possible_values = [None] * nvars
row_solns = []
for row, v in zip(A, V):
lut = []
ws = []
var_assignments = []
for j, val in enumerate(row):
if val:
lut.append(j)
ws.append(val)
var_assignments.append(set())
for soln in WeightedPartitions(ws, v):
for i, w in enumerate(soln):
var_assignments[i].add(w)
for j, assignments in zip(lut, var_assignments):
if possible_values[j] is None:
possible_values[j] = assignments
else:
possible_values[j] &= assignments
return tuple(frozenset(x) for x in possible_values)
def WeightedPartitionSolver(L, B, n):
# the entries of L must be non-negative
# B gives valid values coming from other equations (see VarBound)
def inner(L, B, n):
if n == 0:
yield (0,) * len(L)
elif L:
w, allowed = L[0], B[0]
L, B = L[1:], B[1:]
for i in range(n // w + 1):
if i in allowed:
for tl in inner(L, B, n - w * i):
yield (i,) + tl
return list(inner(L, B, n))
def Filter(PP, X, W):
if [] in PP:
return None
while True:
for Wi, P in zip(W, PP):
F = FixedPoints(P)
for j in F:
P0j = P[0][j]
Wij = Wi[1][j]
if X[Wij] == -1:
X[Wij] = P0j
elif X[Wij] != P0j:
return None
LL=[]
for Wi, P in zip(W, PP):
LL.append([p for p in P if not any(X[idx] not in (-1, pval) for idx, pval in zip(Wi[1], p))])
if not LL[-1]:
return None
if PP == LL:
return LL, X
PP = LL
def FixedPoints(P):
# This would prefer P to be transposed
m=len(P)
n=len(P[0])
return tuple(i for i in range(n) if all(P[j][i] == P[0][i] for j in range(m)))
A simpler brute-force code by Max Alekseyev is available in this answer.

Infeasible solution by pulp [duplicate]

I'm trying to solve an assignment problem with pulp. The basic part of the code is as follows:
set_I = range(1, numberOfPoints)
set_J = range(1, numberOfCentroids)
tau = 0.15
Q = 15
# decision variable
x_vars = LpVariable.dicts(name="x_vars", indexs=(set_I, set_J), lowBound=0, upBound=1, cat=LpInteger)
# model name
prob = LpProblem("MIP_Model", LpMinimize)
# constraints
for i in set_I:
prob += lpSum(x_vars[i][j] for j in set_J) == 1, ""
for j in set_J:
prob += lpSum(x_vars[i][j] for i in set_I) >= 1, ""
for j in set_J:
prob += lpSum(x_vars[i][j] for i in set_I) <= Q*(1-tau), ""
for j in set_J:
prob += lpSum(x_vars[i][j] for i in set_I) >= Q*(1+tau), ""
# objective
prob += lpSum(d[i, j]*x_vars[i][j] for i in set_I for j in set_J)
prob.solve()
The result is like this:
Problem MODEL has 31 rows, 76 columns and 304 elements
Coin0008I MODEL read with 0 errors
Problem is infeasible - 0.01 seconds
Option for printingOptions changed from normal to all
However, the problem is not infeasible and results are obtained with other solvers.
I wonder if there is a syntax error and is the problem caused by this?
I have asked a similar question in the next link:
Infeasible solution by pulp

When I run the problem locally, with d a matrix of ones, 20 points, and 3 centroids. It also becomes infeasible for me. Look at the constraints:
_C22: x_vars_10_1 + x_vars_11_1 + x_vars_12_1 + x_vars_13_1 + x_vars_14_1
+ x_vars_15_1 + x_vars_16_1 + x_vars_17_1 + x_vars_18_1 + x_vars_19_1
+ x_vars_1_1 + x_vars_2_1 + x_vars_3_1 + x_vars_4_1 + x_vars_5_1 + x_vars_6_1
+ x_vars_7_1 + x_vars_8_1 + x_vars_9_1 <= 12.75
_C23: x_vars_10_2 + x_vars_11_2 + x_vars_12_2 + x_vars_13_2 + x_vars_14_2
+ x_vars_15_2 + x_vars_16_2 + x_vars_17_2 + x_vars_18_2 + x_vars_19_2
+ x_vars_1_2 + x_vars_2_2 + x_vars_3_2 + x_vars_4_2 + x_vars_5_2 + x_vars_6_2
+ x_vars_7_2 + x_vars_8_2 + x_vars_9_2 <= 12.75
_C24: x_vars_10_1 + x_vars_11_1 + x_vars_12_1 + x_vars_13_1 + x_vars_14_1
+ x_vars_15_1 + x_vars_16_1 + x_vars_17_1 + x_vars_18_1 + x_vars_19_1
+ x_vars_1_1 + x_vars_2_1 + x_vars_3_1 + x_vars_4_1 + x_vars_5_1 + x_vars_6_1
+ x_vars_7_1 + x_vars_8_1 + x_vars_9_1 >= 17.25
_C25: x_vars_10_2 + x_vars_11_2 + x_vars_12_2 + x_vars_13_2 + x_vars_14_2
+ x_vars_15_2 + x_vars_16_2 + x_vars_17_2 + x_vars_18_2 + x_vars_19_2
+ x_vars_1_2 + x_vars_2_2 + x_vars_3_2 + x_vars_4_2 + x_vars_5_2 + x_vars_6_2
+ x_vars_7_2 + x_vars_8_2 + x_vars_9_2 >= 17.25
You require
x_vars_10_2 + x_vars_11_2 + x_vars_12_2 + x_vars_13_2 + x_vars_14_2
+ x_vars_15_2 + x_vars_16_2 + x_vars_17_2 + x_vars_18_2 + x_vars_19_2
+ x_vars_1_2 + x_vars_2_2 + x_vars_3_2 + x_vars_4_2 + x_vars_5_2 + x_vars_6_2
+ x_vars_7_2 + x_vars_8_2 + x_vars_9_2
to be greater than 17.25 and smaller than 12.75 at the same time. That's not possible, of course.

numpy.ndarray lambda is not callable within loop when using scipy.minimize_scalar, works fine single-run

I'm working on a piece of code that calculates surface wave velocities. When I run this as a single-shot in my Jupyter notebook it works fine, however when I try to start looping over the phi1 angle, I end up with the 'numpy.ndarray is not callable error' after the zero iteration of the loop.
It seems to be focused on the lambda (g) I am using, but I am not sure the best way to go about debugging it. Any help/review would be greatly appreciated as Google searches aren't helping too much.
I do apologize for the novel of code, but this is trimmed down to where it should be easily reproducible.
## This code is an implementation of the framework described by M. Cherry et al. in
## Cherry, Matthew R., Shamachary Sathish, and Ramana Grandhi. "A numerical method for predicting Rayleigh surface wave
## velocity in anisotropic crystals." Journal of Computational Physics 351 (2017): 108-120.
## The logic of this code is as follows.
## (1) It then calculates R, T, and Qv (which is the velocity independent Q, needed to determine the velocity range
## over which to search.
## (2) It then iterates to determine the velocity range by changing a variable called 'angle'.
## (3) It then finds a minimum vL, and solves for Q (with velocity as a term), N, A, B, and Z matrices.
## (4) It stores the three Euler angles (p1, P, p2) and RSW_vel
## This does not deal with pseudowaves yet.
import matplotlib.pyplot as plt
import numpy as np
import scipy as sp
import scipy.optimize as spo
import scipy.linalg as la
from scipy.linalg import eig, eigvals
import math
from decimal import *
import time
ts1 = time.time()
# Setup material properties
# Setup the Cij matrix of Ni (units in Pa)
C11 = C22 = C33 = 2.41e11
C12 = C13 = C23 = C21 = C31 = C32 = 1.47e11
C14 = C15 = C16 = C24 = C25 = C26 = C34 = C35 = C36 = C41 = C42 = C43 = C51 = C52 = C53 = C61 = C62 = C63 = 0
C44 = C55 = C66 = 1.26e11
C45 = C46 = C56 = C54 = C64 = C65 = 0
Cij_list = np.array([C11,C12,C13,C14,C15,C16,C21,C22,C23,C24,C25,C26,C31,C32,C33,C34,C35,C36,C41,C42,C43,C44,C45,C46,C51,C52,C53,C54,C55,C56,C61,C62,C63,C64,C65,C66])
Cij=Cij_list.reshape(6,6)
print('Cij =')
print(Cij)
# Setup density of Ni (units are in kg/m3)
rho = 8910
RSW_matrix = np.array([0,0,0,0])
## This is where the setting up of the loop for p1, P, and p2 should be.
# Setup Euler Angles in Bunge's active notation (one time only - demo, will need to be iterated over), angles in degrees. p1 = phi1, P = Phi, p2 = phi2
#p1 = 15
P = 45
p2 = 8
for rotation in range(0, 180):
## Below this point, everything must be done for every Euler calculation.
p1_rad = np.radians(rotation)
P_rad = np.radians(P)
p2_rad = np.radians(p2)
Eul_Ang_deg = np.array([p1,P,p2])
Eul_Ang_rad = np.array([p1_rad,P_rad,p2_rad])
c1 = np.cos(p1_rad)
c2 = np.cos(P_rad)
c3 = np.cos(p2_rad)
s1 = np.sin(p1_rad)
s2 = np.sin(P_rad)
s3 = np.sin(p2_rad)
# Solve Rotation Matrix from Bunge's active notation, and m and n (m is wave propagation vector, n is plane normal)
Rot_mat = np.array([[c1*c3-c2*s1*s3,-c1*s3-c2*c3*s1,s1*s2],
[c3*s1+c1*c2*s3,c1*c2*c3-s1*s3,-c1*s2],
[s2*s3,c3*s2,c2]]) #- could fill in the rest here and just extract col 1 and col 3
print('Rotation Matrix =')
print(Rot_mat)
m_vec = Rot_mat[[0],:]
m1 = Rot_mat[[0],[0]]
m2 = Rot_mat[[0],[1]]
m3 = Rot_mat[[0],[2]]
n_vec = Rot_mat[[2],:]
n1 = Rot_mat[[2],[0]]
n2 = Rot_mat[[2],[1]]
n3 = Rot_mat[[2],[2]]
# Below, we calculate R, T and Q matrices
# CORRECT R matrix calculations
Rij = np.ndarray(shape=(3,3))
Rij[0,0] = C11*m1*n1 + C61*m2*n1 + C51*m3*n1 + C16*m1*n2 + C66*m2*n2 + C56*m3*n2 + C15*m1*n3 + C65*m2*n3 + C55*m3*n3
Rij[0,1] = C16*m1*n1 + C66*m2*n1 + C56*m3*n1 + C12*m1*n2 + C62*m2*n2 + C52*m3*n2 + C14*m1*n3 + C64*m2*n3 + C54*m3*n3
Rij[0,2] = C15*m1*n1 + C65*m2*n1 + C55*m3*n1 + C14*m1*n2 + C64*m2*n2 + C54*m3*n2 + C13*m1*n3 + C63*m2*n3 + C53*m3*n3
Rij[1,0] = C61*m1*n1 + C21*m2*n1 + C41*m3*n1 + C66*m1*n2 + C26*m2*n2 + C46*m3*n2 + C65*m1*n3 + C25*m2*n3 + C45*m3*n3
Rij[1,1] = C66*m1*n1 + C26*m2*n1 + C46*m3*n1 + C62*m1*n2 + C22*m2*n2 + C42*m3*n2 + C64*m1*n3 + C24*m2*n3 + C44*m3*n3
Rij[1,2] = C65*m1*n1 + C25*m2*n1 + C45*m3*n1 + C64*m1*n2 + C24*m2*n2 + C44*m3*n2 + C63*m1*n3 + C23*m2*n3 + C43*m3*n3
Rij[2,0] = C51*m1*n1 + C41*m2*n1 + C31*m3*n1 + C56*m1*n2 + C46*m2*n2 + C36*m3*n2 + C55*m1*n3 + C45*m2*n3 + C35*m3*n3
Rij[2,1] = C56*m1*n1 + C46*m2*n1 + C36*m3*n1 + C52*m1*n2 + C42*m2*n2 + C32*m3*n2 + C54*m1*n3 + C44*m2*n3 + C34*m3*n3
Rij[2,2] = C55*m1*n1 + C45*m2*n1 + C35*m3*n1 + C54*m1*n2 + C44*m2*n2 + C34*m3*n2 + C53*m1*n3 + C43*m2*n3 + C33*m3*n3
Rij_trans = np.transpose(Rij)
Rij_inv = np.linalg.inv(Rij)
# CORRECT T matrix calculations
Tij = np.ndarray(shape=(3,3))
Tij[0,0] = C11*n1*n1 + C16*n2*n1 + C51*n3*n1 + C16*n1*n2 + C66*n2*n2 + C56*n3*n2 + C15*n1*n3 + C65*n2*n3 + C55*n3*n3
Tij[0,1] = C16*n1*n1 + C66*n2*n1 + C56*n3*n1 + C12*n1*n2 + C62*n2*n2 + C52*n3*n2 + C14*n1*n3 + C64*n2*n3 + C54*n3*n3
Tij[0,2] = C15*n1*n1 + C65*n2*n1 + C55*n3*n1 + C14*n1*n2 + C64*n2*n2 + C54*n3*n2 + C13*n1*n3 + C63*n2*n3 + C53*n3*n3
Tij[1,0] = C61*n1*n1 + C21*n2*n1 + C41*n3*n1 + C66*n1*n2 + C26*n2*n2 + C46*n3*n2 + C65*n1*n3 + C25*n2*n3 + C45*n3*n3
Tij[1,1] = C66*n1*n1 + C26*n2*n1 + C46*n3*n1 + C62*n1*n2 + C22*n2*n2 + C42*n3*n2 + C64*n1*n3 + C24*n2*n3 + C44*n3*n3
Tij[1,2] = C65*n1*n1 + C25*n2*n1 + C45*n3*n1 + C64*n1*n2 + C24*n2*n2 + C44*n3*n2 + C63*n1*n3 + C23*n2*n3 + C43*n3*n3
Tij[2,0] = C51*n1*n1 + C41*n2*n1 + C31*n3*n1 + C56*n1*n2 + C46*n2*n2 + C36*n3*n2 + C55*n1*n3 + C45*n2*n3 + C35*n3*n3
Tij[2,1] = C56*n1*n1 + C46*n2*n1 + C36*n3*n1 + C52*n1*n2 + C42*n2*n2 + C32*n3*n2 + C54*n1*n3 + C44*n2*n3 + C34*n3*n3
Tij[2,2] = C55*n1*n1 + C45*n2*n1 + C35*n3*n1 + C54*n1*n2 + C44*n2*n2 + C34*n3*n2 + C53*n1*n3 + C43*n2*n3 + C33*n3*n3
Tij_trans = np.transpose(Tij)
Tij_inv = np.linalg.inv(Tij)
## R and T defined as above. Q needs to be new - and not remove the rv2. Thus, we introduce Qv for "Q to determine
## critical velocity"
Qv = np.ndarray(shape=(3,3))
Qv[0,0] = C11*m1*m1 + C61*m2*m1 + C51*m3*m1 + C16*m1*m2 + C66*m2*m2 + C56*m3*m2 + C15*m1*m3 + C65*m2*m3 + C55*m3*m3
Qv[0,1] = C16*m1*m1 + C66*m2*m1 + C56*m3*m1 + C12*m1*m2 + C62*m2*m2 + C52*m3*m2 + C14*m1*m3 + C64*m2*m3 + C54*m3*m3
Qv[0,2] = C15*m1*m1 + C65*m2*m1 + C55*m3*m1 + C14*m1*m2 + C64*m2*m2 + C54*m3*m2 + C13*m1*m3 + C63*m2*m3 + C53*m3*m3
Qv[1,0] = C61*m1*m1 + C21*m2*m1 + C41*m3*m1 + C66*m1*m2 + C26*m2*m2 + C46*m3*m2 + C65*m1*m3 + C25*m2*m3 + C45*m3*m3
Qv[1,1] = C66*m1*m1 + C26*m2*m1 + C46*m3*m1 + C62*m1*m2 + C22*m2*m2 + C42*m3*m2 + C64*m1*m3 + C24*m2*m3 + C44*m3*m3
Qv[1,2] = C65*m1*m1 + C25*m2*m1 + C45*m3*m1 + C64*m1*m2 + C24*m2*m2 + C44*m3*m2 + C63*m1*m3 + C23*m2*m3 + C43*m3*m3
Qv[2,0] = C51*m1*m1 + C41*m2*m1 + C31*m3*m1 + C56*m1*m2 + C46*m2*m2 + C36*m3*m2 + C55*m1*m3 + C45*m2*m3 + C35*m3*m3
Qv[2,1] = C56*m1*m1 + C46*m2*m1 + C36*m3*m1 + C52*m1*m2 + C42*m2*m2 + C32*m3*m2 + C54*m1*m3 + C44*m2*m3 + C34*m3*m3
Qv[2,2] = C55*m1*m1 + C45*m2*m1 + C35*m3*m1 + C54*m1*m2 + C44*m2*m2 + C34*m3*m2 + C53*m1*m3 + C43*m2*m3 + C33*m3*m3
res = []
res = spo.minimize_scalar(lambda x: ((min(eigvals(Qv * np.cos(np.radians(x)) * np.cos(np.radians(x)) + (Rij + Rij_trans) * np.cos(np.radians(x)) * np.sin(np.radians(x)) + Tij * np.sin(np.radians(x)) * np.sin(np.radians(x)))) / rho)**(0.5))*(1/(np.abs(np.cos(np.radians(x))))), bracket=(-90,0,90), method='Golden')
vel_L = res.fun.real
vel_L_int = int(vel_L)
print('velocity limit')
print(vel_L)
print(vel_L_int)
Det_B_try = 1
Qij = np.zeros(shape=(3,3))
Qij[0,1] = C16*m1*m1 + C66*m2*m1 + C56*m3*m1 + C12*m1*m2 + C62*m2*m2 + C52*m3*m2 + C14*m1*m3 + C64*m2*m3 + C54*m3*m3
Qij[0,2] = C15*m1*m1 + C65*m2*m1 + C55*m3*m1 + C14*m1*m2 + C64*m2*m2 + C54*m3*m2 + C13*m1*m3 + C63*m2*m3 + C53*m3*m3
Qij[1,0] = C61*m1*m1 + C21*m2*m1 + C41*m3*m1 + C66*m1*m2 + C26*m2*m2 + C46*m3*m2 + C65*m1*m3 + C25*m2*m3 + C45*m3*m3
Qij[1,2] = C65*m1*m1 + C25*m2*m1 + C45*m3*m1 + C64*m1*m2 + C24*m2*m2 + C44*m3*m2 + C63*m1*m3 + C23*m2*m3 + C43*m3*m3
Qij[2,0] = C51*m1*m1 + C41*m2*m1 + C31*m3*m1 + C56*m1*m2 + C46*m2*m2 + C36*m3*m2 + C55*m1*m3 + C45*m2*m3 + C35*m3*m3
Qij[2,1] = C56*m1*m1 + C46*m2*m1 + C36*m3*m1 + C52*m1*m2 + C42*m2*m2 + C32*m3*m2 + C54*m1*m3 + C44*m2*m3 + C34*m3*m3
for v in range(1, vel_L_int):
# Calculate rho vel-squared
rv2 = rho * v ** 2
# CORRECT Q matrix calculations
Qij[0,0] = C11*m1*m1 + C61*m2*m1 + C51*m3*m1 + C16*m1*m2 + C66*m2*m2 + C56*m3*m2 + C15*m1*m3 + C65*m2*m3 + C55*m3*m3 - rv2
Qij[1,1] = C66*m1*m1 + C26*m2*m1 + C46*m3*m1 + C62*m1*m2 + C22*m2*m2 + C42*m3*m2 + C64*m1*m3 + C24*m2*m3 + C44*m3*m3 - rv2
Qij[2,2] = C55*m1*m1 + C45*m2*m1 + C35*m3*m1 + C54*m1*m2 + C44*m2*m2 + C34*m3*m2 + C53*m1*m3 + C43*m2*m3 + C33*m3*m3 - rv2
Qij_trans = np.transpose(Qij)
Qij_inv = np.linalg.inv(Qij)
# Create a new Nik matrix
TR_ik = -1*np.matmul(Tij_inv, Rij_trans)
TI_ik = Tij_inv
QRTR_ik = -Qij + np.matmul(np.matmul(Rij, Tij_inv), Rij_trans)
RT_ik = -1*np.matmul(Rij, Tij_inv)
# Nik Creation
Nik_r1to3 = np.concatenate((TR_ik,TI_ik),axis=1)
Nik_r4to6 = np.concatenate((QRTR_ik,RT_ik),axis=1)
Nik_whole = np.concatenate((Nik_r1to3,Nik_r4to6))
# Eigenvalues and Eigenvectors of Nik
eigvals, eigvecs = eig(Nik_whole)
## The following code sorts the eigenvalues for those with negative imaginary numbers, and conducts the matrix
## operations to generate the B, A, and Z matrices from the COLUMNS in Pythons representation of eigenvectors
eigvecsT = np.transpose(eigvecs)
output_ab = output_ab_neg_imag = eigvecsT[eigvals.imag < 0]
output_abT = np.transpose(output_ab)
### This is the B matrix (bottom 3 vectors assumed as b)
B_matrix = output_b = np.delete(output_abT, [0,1,2], axis=0)
B_matrixT = np.transpose(B_matrix)
### This is the A matrix (these are the displacement vectors, and when A is combined with B, the surface impedance (Z) matrix can be found)
A_matrix = output_a = np.delete(output_abT, [3,4,5], axis=0)
det_B = np.linalg.det(B_matrix)
det_cc_B = det_B.conjugate()
det_B_x_det_cc_B = det_B*det_cc_B
v_real = v.real
det_B_x_det_cc_B_real = det_B_x_det_cc_B.real
if det_B_x_det_cc_B_real < Det_B_try:
RSW_vel = v_real
Det_B_try = det_B_x_det_cc_B_real
print('RSW =')
print(RSW_vel)
print('Min |B|x|B|* =')
print(Det_B_try)
col_to_be_added2 = np.array([p1, P, p2, RSW_vel])
RSW_matrix = np.column_stack((RSW_matrix,col_to_be_added2))
print('RSW vel matrix is')
print(RSW_matrix)

IndexError: list assignment index out of range. How do I solve this?

I am trying the Newmark's constant average acceleration method. I am getting this error. How do I recover from this error?
IndexError Traceback (most recent call last)
41 for i in range(len(t)):
42 pn[i+1] = p[i+1]+ a1*u[i] + a2*v[i] + a3*a[i]
43 u[i+1] = pn[i+1]/kn
44 v[i+1] = y*(u[i+1]-u[i])/(b*dt) + (1-y/b)v[i] + dt (1-y/(2*b))*a[i]
IndexError: list assignment index out of range
y = 1/2
b = 1/4
u = []
v = []
t = []
p = [0,25,43.3013,50,43.3013,25,0,0,0,0,0,0]
a = []
pn = []
pn.append(0)
x = 0.0
for i in range(11):
z = 0.0 + x
t.append(z)
x = x + 0.1
m = 0.45594
k = 18
c = 0.2865
u.append(0)
v.append(0)
a.append((p[0]-c*v[0]-k*u[0])/m)
dt = 0.1
a1 =(m/(b*dt*dt)+y*c/(b*dt))
a2 = (m/(b*dt)+(y/b-1)*c)
a3 = (((1/(2*b))-1)*m + dt*((y/(2*b))-1)*c)
kn = k + a1
for i in range(len(t)-1):
pn[i+1] = p[i+1]+ a1*u[i] + a2*v[i] + a3*a[i]
u[i+1] = pn[i+1]/kn
v[i+1] = y*(u[i+1]-u[i])/(b*dt) + (1-y/b)*v[i] + dt* (1-y/(2*b))*a[i]
a[i+1] = (u[i+1]-u[i])/(b*dt*dt) - v[i]/(b*dt)-(1/(2*b)-1)*a[i]

Your pn, a, u, v are defined as a list with length 1, so there is no index such as pn[1]. You can use append or define the list with needed length.
for i in range(len(t)):
pn.append(p[i+1] + a1*u[i] + a2*v[i] + a3*a[i])
u.append(pn[i+1]/kn)
v.append(y*(u[i+1]-u[i])/(b*dt) + (1-y/b)*v[i] + dt* (1-y/(2*b))*a[i])
a.append((u[i+1]-u[i])/(b*dt*dt) - v[i]/(b*dt)-(1/(2*b)-1)*a[i])
or
pn, a, u, v = [0]*11, [0]*11, [0]*11 [0]*11
pn[0], u[0], v[0] = 0, 0, 0
a[0] = (p[0]-c*v[0]-k*u[0])/m
...
for i in range(len(t)-1):
pn[i+1] = p[i+1] + a1*u[i] + a2*v[i] + a3*a[i]
u[i+1] = pn[i+1]/kn
v[i+1] = y*(u[i+1]-u[i])/(b*dt) + (1-y/b)*v[i] + dt* (1-y/(2*b))*a[i]
a[i+1] = (u[i+1]-u[i])/(b*dt*dt) - v[i]/(b*dt)-(1/(2*b)-1)*a[i]

integrate cos(x)*cos(2x)*...*cos(mx) via SAGE

I'm going to find $I_m=\int_0^{2\pi} \prod_{k=1}^m cos(kx){}dx$, where $m=1,2,3\ldots$
Simple SAGE code:
x=var('x')
f = lambda m,x : prod([cos(k*x) for k in range(1,m+1)])
for m in range(1,15+1):
print m, numerical_integral(f(m,x), 0, 2*pi)[0],integrate(f(m,x),x,0,2*pi).n()
Output:
1 -1.47676658757e-16 0.000000000000000
2 -5.27735962315e-16 0.000000000000000
3 1.57079632679 1.57079632679490
4 0.785398163397 0.785398163397448
5 -2.60536121164e-16 0.000000000000000
6 -1.81559273097e-16 0.000000000000000
7 0.392699081699 0.392699081698724
8 0.343611696486 0.147262155637022
9 -1.72448482421e-16 0.294524311274043
10 -1.8747663502e-16 0.196349540849362
11 0.214757310304 0.312932080728671
12 0.190213617698 0.177941771394734
13 -1.30355375996e-16 0.208621387152447
14 -1.25168280013e-16 0.0859029241215959
15 0.138441766107 0.134223318939994
As you can see numerical answer is right, but result of integrate(...) is right for $m=1,2,\ldots,7$ and then there is some bug.
We can print indefinite integral:
for m in range(7,11+1):
print 'm=',m
print 'Indef_I_m=',integrate(f(m,x),x)
And Output:
m = 7
Indef_I_m = 1/16*x + 1/16*sin(2*x) + 1/32*sin(4*x) + 7/384*sin(6*x) +
7/512*sin(8*x) + 3/320*sin(10*x) + 5/768*sin(12*x) + 5/896*sin(14*x) +
1/256*sin(16*x) + 1/384*sin(18*x) + 1/640*sin(20*x) + 1/704*sin(22*x) +
1/1536*sin(24*x) + 1/1664*sin(26*x) + 1/1792*sin(28*x)
m = 8
Indef_I_m = 3/128*x + 5/256*sin(2*x) + 1/32*sin(3*x) + 5/512*sin(4*x) +
5/768*sin(6*x) + 1/256*sin(8*x) + 1/256*sin(10*x) + 1/256*sin(12*x) +
1/256*sin(14*x) + 1/256*sin(16*x) + 7/2304*sin(18*x) + 3/1280*sin(20*x)
+ 5/2816*sin(22*x) + 1/768*sin(24*x) + 3/3328*sin(26*x) +
1/1792*sin(28*x) + 1/1920*sin(30*x) + 1/4096*sin(32*x) +
1/4352*sin(34*x) + 1/4608*sin(36*x) + 3/32*sin(x)
m = 9
Indef_I_m = 3/64*x + 3/128*sin(2*x) + 23/768*sin(3*x) + 3/256*sin(4*x) +
3/640*sin(5*x) + 1/128*sin(6*x) + 5/1792*sin(7*x) + 5/2304*sin(9*x) +
3/2816*sin(11*x) + 1/832*sin(13*x) + 1/1280*sin(15*x) + 3/4352*sin(17*x)
+ 5/4864*sin(19*x) + 1/1344*sin(21*x) + 3/2944*sin(23*x) +
7/6400*sin(25*x) + 1/1152*sin(27*x) + 3/3712*sin(29*x) +
5/7936*sin(31*x) + 1/2112*sin(33*x) + 3/8960*sin(35*x) +
1/4736*sin(37*x) + 1/4992*sin(39*x) + 1/10496*sin(41*x) +
1/11008*sin(43*x) + 1/11520*sin(45*x) + 23/256*sin(x)
m = 10
Indef_I_m = 1/32*x + 1/64*sin(2*x) + 17/512*sin(3*x) + 1/128*sin(4*x) +
7/2560*sin(5*x) + 1/192*sin(6*x) + 3/1792*sin(7*x) + 1/1152*sin(9*x) +
5/5632*sin(11*x) + 3/6656*sin(13*x) + 1/2560*sin(15*x) +
5/8704*sin(17*x) + 3/9728*sin(19*x) + 1/2688*sin(21*x) +
1/2944*sin(23*x) + 1/6400*sin(25*x) + 1/4608*sin(27*x) +
3/14848*sin(29*x) + 3/15872*sin(31*x) + 5/16896*sin(33*x) +
3/8960*sin(35*x) + 3/9472*sin(37*x) + 1/3328*sin(39*x) +
5/20992*sin(41*x) + 1/5504*sin(43*x) + 1/7680*sin(45*x) +
1/12032*sin(47*x) + 1/12544*sin(49*x) + 1/26112*sin(51*x) +
1/27136*sin(53*x) + 1/28160*sin(55*x) + 13/128*sin(x)
m = 11
Indef_I_m = 51/1024*x + 53/2048*sin(2*x) + 13/768*sin(3*x) + 53/4096*sin(4*x) +
13/1536*sin(6*x) + 1/2048*sin(8*x) + 1/2560*sin(10*x) + 1/3072*sin(12*x)
+ 5/14336*sin(14*x) + 1/4096*sin(16*x) + 5/18432*sin(18*x) +
1/4096*sin(20*x) + 1/5632*sin(22*x) + 5/24576*sin(24*x) +
5/26624*sin(26*x) + 5/28672*sin(28*x) + 1/5120*sin(30*x) +
3/16384*sin(32*x) + 5/34816*sin(34*x) + 1/9216*sin(36*x) +
5/38912*sin(38*x) + 1/10240*sin(40*x) + 1/10752*sin(42*x) +
3/22528*sin(44*x) + 3/23552*sin(46*x) + 1/8192*sin(48*x) +
3/25600*sin(50*x) + 5/53248*sin(52*x) + 1/13824*sin(54*x) +
3/57344*sin(56*x) + 1/29696*sin(58*x) + 1/30720*sin(60*x) +
1/63488*sin(62*x) + 1/65536*sin(64*x) + 1/67584*sin(66*x) +
13/256*sin(x)
so for $m=7$ answer is right compare with Indef_I_7 via WolframAlpha
and for $m=8$ answer is incorrect Indef_I_8 via WolframAlpha
There should be Indef_I_8=$\frac{7x}{128}+\ldots$ and no $\sin(x)$, $\sin(3x)$ in summation, only $\sin(2k)$ for $k=1,2,3,\ldots 18$
Sorry for volumetric calculations !
The question is - Am I right that it is the bug in the symbolic integration?

Well, apparently setting algorithm='mathematica_free' solved the issue; this is probably a bug in the default algorithm used bye SAGE ('maxima').

So the answer is - Yes. It is a bug in algorithm='maxima', so use algorithm='mathematica_free' (def new function to find definite integral) or simplify_full() for such product of cos(kx) and than integrate.
f(8,x).simplify_full().integral(x,0,2pi) == 7/64pi and
(7/64*pi).n() == 0.343611696486384 is correct