DESIGN OPTIMIZATION OF POST-TENSIONED SLAB
CONCERNING FLOOR VIBRATION
Adisorn Owatsiriwong
POSTECK PRESTRESSING
Problem Statement:
This simple example demonstrates the design optimization of post-tensioned slab
that concerns to floor vibration. The lower bound of fundamental frequency is
included in design constraints to assure that the slab is not prone to vibration under
human excited activities.
10.0m
10.0m
SDL = 3.0 kPa
LL = 3.0 kPa
2@ 3.0 m
10.0m
COL 0.50x0.50 m
Design Objective: Minimize cost of slab construction
Cost /m2 = 2750 h + 75 Ws (THB / m2)
h = slab thickness (m)
Ws = strand weight per area (kg/m2)
1|Page
To simplify the formulation, the design variable shall be changed from Ws (strand
density) to F/A. The nonlinear cost function can be rewritten as:
Ws = 4 N/L = 2(F/A) x h/110x103
Cost/m2 = 2750 h + 150 (F/A) h/110x103
Where, N = Number of strand per design strip
L = panel span length (m)
Subjected to:
1. First mode eigen frequency f1 > 3.5 Hz
2. Tensile stress at bottom of slab is less than 0.5sqrt(fc’) at service load (ACI
318-11) with SDL = 3.0 kPa, LL = 3.0 kPa
3. Slab thickness shall be less than 350 mm
4. Average precompressive stress shall be in range of [1.2, 2.5] MPa
5. Percentage of balance load shall be between 75-125% of slab selfweight
Since the slab has the same dimension in both directions, we only perform
analysis in one direction.
1. Using finite element analysis, the 1st mode eigen frequency can be obtained
from the empirical expression:
f1 = 17.8 h - 0.25 Hz ; h = slab thickness (m)
2|Page
Figure 1. Modal test of slabs with various thickness using ETABS
2. Compute Tensile stress at mid span:
ft = -F/A + Mnet/S
Mnet = (wtot-wbal) L2/8
Wbal = 8(F/A)x A a/L2
a = 0.5h – 0.05
A = 0.5Lh
3. h < 0.35 m
4. 1.1 MPa < F/A < 2.5 MPa
5. % balance load between 70% to 125% of slab self weight
Using MATLAB optimization toolbox, we can define the cost function and
nonlinear constraint functions as followings:
function out = Cost(x)
% Cost function
Ws = 2.0*x(2)*x(1)/110e3;
out = 2750.0*x(1) + 75*Ws;
end
3|Page
function [c,ceq] = nlcon(x)
L = 10.0; % span length
A = 0.5*L*x(1);
S = 0.5*L*x(1)^2/6.;
w = (24*x(1)+6.0)*0.5*L; % SDL = 3.0, LL = 3.0
a = 0.5*x(1)-0.05;
wbal = 8*x(2)*1e3*a*A/L^2;
Mnet = (w-wbal)*L^2/8.;
c(1) = -x(2) + Mnet/S*1e-3-2.80; % ft < 0.5sqrt(fc')
c(2) = -17.8*x(1) + 0.25 - 3.5; % f1 >= 3.5 Hz
c(3) = wbal-1.25*24*x(1)*0.5*L;
c(4) = -wbal+0.75*24*x(1)*0.5*L;
ceq = []
end
Nonlinear constrained optimization function using fmincon is listed as following:
function [x,fval,exitflag,output,lambda,grad,hessian] = opti_slab(x0,lb,ub)
%% Start with the default options
options = optimoptions('fmincon');
%% Modify options setting
options = optimoptions(options,'Display', 'off');
options = optimoptions(options,'PlotFcns', { @optimplotx @optimplotfval });
[x,fval,exitflag,output,lambda,grad,hessian] = ...
fmincon(@Cost,x0,[],[],[],[],lb,ub,@nlcon,options);
Figure 2 Optimization results
4|Page
After 6th iteration, the optimal point is found to be [0.312, 2.5] (slab thickness =
0.312 m and F/A = 2.5 MPa).
The final design slab is verified by PTDATA (Seneca Structural Engineering Inc.) to
assure that stress, deflection and strength of slab are satisfactory.
Figure Design verification using PTDATA
5|Page
Conclusion
In this example it is noted that the optimal solution stops when F/A reaches to
upper limit. This might because the cost/efficiency of post-tensioning material is
quite low as compared with the unit price of concrete. It is possible to consider the
larger limit of F/A for the better solution. However, this might cause another
problem due to larger slab shortening and long term performance of slab.
6|Page
