We have product which works on three phase AC supply. We were facing issue on I2C interface of external RTC due to noise from external environment. After some inspection, we realized that because of improper isolation of analog and digital ground, I2C device was getting reset due to bouncing of ground level. We eliminated that issue to some extent by isolating ground. Still the issue persists with I2C interface of external RTC on noisy sites. We obviously never faced noise issue in our lab, so I want to know how we can generate noise in three phase supply going to the device so that we can replicate the issue in the lab?
Ask the lab what they use and buy one for yourself, keep in mind test equipment is expensive. A better thing to do is use proper grounding and shielding practices.
If the EMI is conducted (through the supply of the device, then you should be able to use line filters, ferrites and capacitors to block the transients from coming in on the supply.
If the EMI is radiating, then use proper shielding to block the radiated emissions from reaching the device.
I suspect you have a ground loop and an improper ground or ground loop. Split grounds are one of the quickest ways to have trouble in a design. It is important to control ground currents in any design and where they end up, this is best accomplished through proper layout.
Get the book Electromagnetic Compatibility Engineering by Henery Ott. It is a really good resource.