tug/examples/Rcpp-interface.R
2022-08-12 14:53:27 +02:00

160 lines
3.4 KiB
R

## Time-stamp: "Last modified 2022-03-16 14:01:11 delucia"
library(Rcpp)
library(RcppEigen)
library(ReacTran)
library(deSolve)
options(width=110)
setwd("app")
## This creates the "diff1D" function with our BTCSdiffusion code
sourceCpp("Rcpp-BTCS-1d.cpp")
### FTCS explicit (same name)
sourceCpp("RcppFTCS.cpp")
## Grid 101
## Set initial conditions
N <- 1001
D.coeff <- 1E-3
C0 <- 1 ## Initial concentration (mg/L)
X0 <- 0 ## Location of initial concentration (m)
## Yini <- c(C0, rep(0,N-1))
## Ode1d solution
xgrid <- setup.grid.1D(x.up = 0, x.down = 1, N = N)
x <- xgrid$x.mid
Diffusion <- function (t, Y, parms){
tran <- tran.1D(C = Y, C.up = 0, C.down = 0, D = parms$D, dx = xgrid)
return(list(tran$dC))
}
## gaussian pulse as initial condition
sigma <- 0.02
Yini <- 0.5*exp(-0.5*((x-1/2.0)**2)/sigma**2)
## plot(x, Yini, type="l")
parms1 <- list(D=D.coeff)
# 1 timestep, 10 s
times <- seq(from = 0, to = 1, by = 0.1)
system.time({
out1 <- ode.1D(y = Yini, times = times, func = Diffusion,
parms = parms1, dimens = N)[11,-1]
})
## Now with BTCS
alpha <- rep(D.coeff, N)
system.time({
out2 <- diff1D(n=N, length=1, field=Yini, alpha=alpha, timestep = 0.1, 0, 0, iterations = 10)
})
## plot(out1, out2)
## abline(0,1)
## matplot(cbind(out1,out2),type="l", col=c("black","red"),lty="solid", lwd=2,
## xlab="grid element", ylab="Concentration", las=1)
## legend("topright", c("ReacTran ode1D", "BTCS 1d"), text.col=c("black","red"), bty = "n")
system.time({
out3 <- RcppFTCS(n=N, length=1, field=Yini, alpha=1E-3, bc_left = 0, bc_right = 0, timestep = 1)
})
## Poor man's
mm <- colMeans(rbind(out2,out3))
matplot(cbind(Yini,out1, out2, out3, mm),type="l", col=c("grey","black","red","blue","green4"), lty="solid", lwd=2,
xlab="grid element", ylab="Concentration", las=1)
legend("topright", c("init","ReacTran ode1D", "BTCS 1d", "FTCS", "poor man's CN"), text.col=c("grey","black","red","blue","green4"), bty = "n")
sum(Yini)
sum(out1)
sum(out2)
sum(out3)
sum(mm)
## Yini <- 0.2*sin(pi/0.1*x)+0.2
## plot(Yini)
## plot(out3)
Fun <- function(dx) {
tmp <- diff1D(n=N, length=1, field=Yini, alpha=alpha, timestep = dx, 0, 0, iterations = floor(1/dx))
sqrt(sum({out1-tmp}^2))
}
reso <- optimise(f=Fun, interval=c(1E-5, 1E-1), maximum = FALSE)
dx <- 0.0006038284
floor(1/dx)
1/dx
system.time({
out2o <- diff1D(n=N, length=1, field=Yini, alpha=alpha, timestep = dx, 0, 0, iterations = 1656)
})
matplot(cbind(out1, out2o),type="l", col=c("black","red"), lty="solid", lwd=2,
xlab="grid element", ylab="Concentration", las=1)
legend("topright", c("ReacTran ode1D", "BTCS 1d dx=0.0006"), text.col=c("black","red"), bty = "n")
dx <- 0.05
system.time({
out2o <- diff1D(n=N, length=1, field=Yini, alpha=alpha, timestep = dx, 0, 0, iterations = 1/dx)
})
matplot(cbind(out1, out2o),type="l", col=c("black","red"), lty="solid", lwd=2,
xlab="grid element", ylab="Concentration", las=1)
legend("topright", c("ReacTran ode1D", "BTCS 1d dx=0.0006"), text.col=c("black","red"), bty = "n")
Matplot
## This creates the "diff1D" function with our BTCSdiffusion code
sourceCpp("Rcpp-BTCS-2d.cpp")
n <- 256
a2d <- rep(1E-3, n^2)
init2d <- readRDS("gs1.rds")
ll <- {init2d - min(init2d)}/diff(range(init2d))
system.time({
res1 <- diff2D(nx=N, ny=N, lenx=1, leny=1, field=ll, alpha=a2d, timestep = 0.1, iterations = 10)
})
hist(ll,32)