Download acoustic_core.py from soundai2016/ADComp: direct link, hf CLI and curl.
- Browser
- Download file 2.37 kB
-
https://huggingface.co/soundai2016/ADComp/resolve/main/acoustic_core.py
- Command line
-
hf download hf://soundai2016/ADComp/acoustic_core.py
-
curl -L -o acoustic_core.py https://huggingface.co/soundai2016/ADComp/resolve/main/acoustic_core.py
2.37 kB
| """Compatible finite-volume operators and physical acoustic ports.""" | |
| import numpy as np | |
| from scipy import sparse | |
| MATERIALS = ['rotated_stripes', 'double_inclusion', 'smooth_checker'] | |
| CONFIG = { | |
| 'source_centres': [[.19,.76],[.54,.47],[.81,.22]], | |
| 'receiver_centres': [[.22,.21],[.79,.77],[.28,.69],[.68,.33]], | |
| 'sigma': .07, 'time_end': 8., 'samples': 641, 'threads': 2, | |
| } | |
| def material(n, name): | |
| x = (np.arange(n)+.5)/n | |
| X,Y = np.meshgrid(x,x,indexing='ij') | |
| if name == 'homogeneous': | |
| rho = np.ones_like(X); bulk = np.ones_like(X) | |
| elif name == 'rotated_stripes': | |
| rho = np.exp(.65*np.sin(4*np.pi*(.8*X+.6*Y))+.1*np.cos(2*np.pi*Y)) | |
| bulk = np.exp(-.45*np.sin(4*np.pi*(.8*X+.6*Y))+.12*np.sin(2*np.pi*X)) | |
| elif name == 'double_inclusion': | |
| a = np.exp(-((X-.34)**2+(Y-.35)**2)/(2*.1**2)) | |
| b = np.exp(-((X-.70)**2+(Y-.67)**2)/(2*.12**2)) | |
| rho = np.exp(1.2*a-.55*b+.1*np.cos(2*np.pi*X)) | |
| bulk = np.exp(-.8*a+.9*b+.1*np.sin(2*np.pi*Y)) | |
| elif name == 'smooth_checker': | |
| z = np.cos(4*np.pi*X)*np.cos(4*np.pi*Y) | |
| rho = np.exp(.8*z+.1*np.sin(2*np.pi*X)) | |
| bulk = np.exp(-.7*z+.15*np.cos(2*np.pi*Y)) | |
| else: | |
| raise ValueError(name) | |
| return X,Y,rho,bulk | |
| def build_sparse(n, name): | |
| X,Y,rho,bulk = material(n,name) | |
| idx = np.arange(n*n).reshape(n,n) | |
| a = np.concatenate([idx[:-1,:].ravel(),idx[:,:-1].ravel()]) | |
| b = np.concatenate([idx[1:,:].ravel(),idx[:,1:].ravel()]) | |
| faces = np.arange(len(a)) | |
| face_rho = .5*(rho.ravel()[a]+rho.ravel()[b]) | |
| D = sparse.coo_matrix((np.concatenate([np.full(len(a),n),np.full(len(a),-n)]), | |
| (np.concatenate([a,b]),np.concatenate([faces,faces]))), | |
| shape=(n*n,len(a))).tocsr() | |
| mass = (1/(n*n*bulk)).ravel() | |
| # K_h=h^2 D rho_face^-1 D^T, A=M^-1/2 K_h M^-1/2=C C^T. | |
| C = sparse.diags(1/np.sqrt(mass)) @ (D/n) @ sparse.diags(1/np.sqrt(face_rho)) | |
| J = sparse.bmat([[None,-C],[C.T,None]],format='csr') | |
| return X,Y,rho,bulk,mass,C,J | |
| def ports(X,Y,bulk,centres,zero_mean): | |
| n = X.shape[0] | |
| columns = [] | |
| for x,y in centres: | |
| f = np.exp(-((X-x)**2+(Y-y)**2)/(2*CONFIG['sigma']**2)) | |
| f *= n*n/f.sum() | |
| if zero_mean: | |
| f -= f.mean() | |
| columns.append((np.sqrt(bulk)*f/n).ravel()) | |
| return np.array(columns).T | |