Hi makarand,
I had faced similar problem years ago. Here are certain points which will guide you to solve the problem:

1. Connect the RC. (R=10 Ohms 2Watts, C=100kpf/400V Make:"EL-CI-AR", available commonly in Mumbai, India". You have not mentioned your location in your user profile. And since your name sounds like you are around Mumbai, so I am assuming so). Make a series RC network, and put it across the terminals of the contactor. If required connect 2 or 3 parallel networks. Connect one such network between the contacts "NO" and "C" of the Relay. Mostly it will reduce the problem. Here is the theory behind it:
When the Contactor is turned ON, the current passes through the Contactor coil. The current stabilizes, WHen you turn OFF the relay, the current through the contactor stops. The contactor coil is an inductor, so it will oppose the change in this current; in doing so it will generate a high reverse voltage. This voltage is generated at the point where the circuit gets opened, i.e. the "C","NO" contacts of the relay. So you see some sparks at the relay contacts typically when relay turns OFF, this will also apply when the relay turning OFF to ON.
To stop this sparking, you have to reduce the current gradually, for which we provide the RC network. So the charge in the inductor finds a path through the RC network to discharge. It depends upon the specifications of the contactor and the quality of the contactor, which I suppose you do not have access to, hence I suggested to put multiple RC networks to reduce the problem in case if One RC network does not solve it.
The theory I gave is very crude explaination, you will find very neat explainations on replies that you will get from other members of this forum typically watch for Kai's replies.
You will find the classic example of how to design products to solve noise problems in the thread "How Good is this Board" posted by Prahlad J. Purohit. Search for it on this forum and you will find the thread.

2. The Wiring: The relay output wires going towards the contactors must not be in close vicinity of the Mains Supply that is given to your instrument. They must be as far as possible. The further they are the lesser will be the problem. And remember they must never run parallel to each other. Treat the mains supply to your instrument as a very sensitive wire. The same rule applies for other wires connected to your instrument, typeically the sensor wires, and any other wires that carry the signal that is connected to the VCC,GND of your CPU.

3. The PCB trace route of the Relay circuit. The PCB trace of the "C", "NO" wire must be very short (length wise) as possible, and must be as thick as possible (widhth wise). This will reduce the resistance and the effect of voltage being developed across the PCB trace which might cause EMI problems with other components/tracks in this PCB. If possible, scratch the green mask OFF the PCB trace, and put extra solder to make the PCB track as thick as possible. Remember to isolate it since it carries high voltages.

4. Look out for any points in your PCBs (all of them), for any component that has loose soldering. Yes, this also gives problems under such situations. I remember about 8 years back one of the instruments was giving a problem and the culprit was: A capacitor that was not soldered. The 7805 had 220uF cap at output, but during the PCB design we had given a provision of another cap 0.1uF disk. But the disk cap was not soldered. The Ground track was given to these caps, and then it was given to the CPU board. Since the cap was not soldered, the ground track had a nice hole of 0.8mm in the 1.5mm thick ground track. We just put the solder on the unsoldered pads and the problem was solved. So watch for such silly mistakes in your product.

Well these are just some of the points that popped up as soon as I read the Noise Problem.


Well thanks for reading the long post.
Bye,
Mr. Kiran V. Sutar.