The time delay between desired input and output is cannot be removed completely with PID controller.  If you try to minimize manually, probably system get more close to an unstable controlling. You have to define the expectations from controller, and optimizing it based on these. Search PID optimizations efforts, like:

http://sts.bwk.tue.nl/7y500/readers/.%5CInstellingenRegelaars_ExtraStof.pdf

http://cdn.intechopen.com/pdfs-wm/45672.pdf

If you do not satisfied with PID performance, you have other possibilities, like Fuzzy-PI controllers. 

https://www.google.rs/webhp?sourceid=chrome-instant&ion=1&espv=2&ie=UTF-8#newwindow=1&q=Fuzzy-PI+controller

Istvan

Is it possible to give some more details with a sketch to better respond e.g

1. what is the input  and output . Let us view the heat exchanger as

 a plant ( say   k1/(k2+s)    ) then assuming a feedback 

what is the input to be varied , its range etc .

2. Suggest a modeling in Pspice where I think you can quickly see

what is happening and also tune it. Also it is better to set up the model

on a per unit basis.

Good luck

Cheers

Thank you so much @  Istvan Kecskes....I Will Try to get best result...Thanks for your information...

Than you @Narasim Ramesh for your support :- I am sending you the transfer function of my system...It is a first order,(By the way it should not be a first order system but according to my knowledge ) but i found the T.F as=(1.96*e(-28.12s))/58.09s+1 , (2.54*e(-42.29s))/55.32s+1  ,  (2.75*e(-35.8s))/49.66s+1.....I have taken 3 different readings and found to be three different T.F. .........So what i supposed to do???

I am also sending the SCADA pic of this system...It Is a SISO system(Just controlling the outlet  temperature of cold water, which gets heated due to hot water heat, flowing through shell and tube type heat exchanger)....

Thanks Tushar for the data .

Some suggestions:

1Begin with a picture of the physics of the problem. Here the hot water

inputs x kw ( as mass*specific heat) flow of heat power(eqt as current I) . This travels across the Heat exchanger to the cold water through thermal resistances(deg c/ watt) (eqt elec resistor ) and masses to be heated up (sp heat *mass *delt theta) (electr a

capacitor) .The voltage correspond to temperature. in eqt circuit. so the syytem

for forced heating is represented electrically as a ladder network of R and C.

2. The above brief analysis leads us to the time delay transfer function of

the system given by cascaded 1/ (RCs + 1)  laplace transfer function. In the

given case it seems we can approximate this with a single 1/ (as + 1) component.

thus a=time constant of heat exchanger . We can set a=(58.09+55.32 +49.66)/3..

.similarly we can average the amplification . Next we are left with e^(-28.12s)

type . Recall this is nothing but f(t-T)u(t-T) type of response . This seems to be

a fixed time delay depending on valve operation etc.In the modeling care has to be taken to incorporate this. It cannot be compensated by the controller and might be the

cause of instability you report.

3. it should be possible to simulate the PID controller as k1 +k2/s+k3s using R, C and L  and plant also as another  Rand C with time delay  Td (maybe tricky to simulate).In any case try without Td first.

4. Dimensional units are important for correlation.

Good Luck

 Cheers

 5 If a different type of control is allowed , the following intuitive solution may be

considered. The aim is to heat up cold water as quickly as possible so turn up

the hot water rate (HWR) to highest temp and mass flow. monitor the cold water temp

and reduce HWR when desired temp is reduced then use simple proportional

controller to maintain this temp.

Hope you find it useful.

Cheers

Thanks for your suggestion @Narasim Ramesh...

There are many types  of PID controller tuning except the conventional one (Zigler & Nicoles), the intelligent types are (Swarm Optimization Techniques) and (Genetic Algorithms), these types used algorithms to search for optimal parameters of PID depending on some cost function.