Empirical "backward" galaxy evolution models for IR bright galaxies are con
strained using multiband IR surveys. A new Monte Carlo algorithm is develop
ed for this task. It exploits a large library of realistic spectral energy
distributions (SEDs) of 837 local IR galaxies (IRAS 25 mum selected) from t
he UV (1000 Angstrom) to the radio (20 cm), including Infrared Space Observ
atory (ISO) measured 3-13 mum unidentified broad features (UIBs). The basic
assumption is that the local correlation between SEDs and mid-infrared (MI
R) luminosities can be applied to earlier epochs of the universe, an assump
tion that will be strongly tested by SIRT F. By attaching an SED appropriat
ely drawn from the SED library to every source predicted by a given model,
the algorithm enables simultaneous comparisons with multiple surveys in a w
ide range of wave bands. Three populations of IR sources are considered in
the evolution models. These include (1) starburst galaxies, (2) normal late
-type galaxies, and (3) galaxies with active galactic nuclei (AGNs). Constr
ained by data from the literature, our best-Dt model ("peak model II) predi
cts that since z = 1.5 the population of starburst galaxies undergoes a ver
y strong luminosity evolution [L = L-0(1 + z)(4.2)] and also strong density
evolution [rho = rho (0)(1 + z)(2)], the normal late-type galaxy populatio
n undergoes a passive luminosity evolution [L = L-0(1 + z)(1.5)], and the g
alaxies with an AGN undergo a pure luminosity evolution similar to that of
optical QSOs [L = L-0(1 + z)(3.5)]. Prior at z greater than or equal to 1.5
all evolution rates drop as (1 + z)(-3). The luminosity evolution results
in evolution of SEDs of IR bright sources because of the luminosity depende
nce of the SEDs. Predictions for number counts, confusion limits, redshift
distributions, and color-color diagrams are made for multiband surveys usin
g the upcoming SIRT F satellite. A Lambda cosmology (Omega (Lambda) = 0.7,
Omega (m) = 0.3, H-0 = 75 km s(-1) Mpc(-1)) is assumed throughout the paper
.