Profinet loosing the connection with Robot during backup from workvisual

your issue is with ProfiNet and you don't even mention installed version....

please read READ FIRST

Only time I've seen ProfiNet drop out due to other non-realtime network traffic was when the network switches were badly configured, and didn't keep low-priority and high-priority traffic separate.

Why is the RPI set to 1ms? Interpolation time is 12ms. I dont have manual to check limitations but if I recall correctly RPI need to be 8ms or more (as it only supports PNET2.25 I think). On KSS 8.3 higher PNET standard is supported (2.31+) so this setting can go down to 1ms... but don’t take my word for, check manual....

Low- and high-priority traffic is not a WorkVisual thing, it's a general network thing. I should have said "deterministic" and "non-deterministic", perhaps.

File transfers, web page requests, are examples of low-priority, non-deterministic traffic. If a few packets get delayed, or even lost, en route, it's no big deal. They use TCP/IP in such a way that they can deal with very noisy and over-crowded network conditions without failing. They simply slow down.

ProfiNet, and other industrial FieldBusses that use Ethernet, are deterministic and hard-realtime, being mission-critical items. They cannot tolerate more than a very small number of packet delays or losses. To be safe, too many delays or losses will cause the driver to fail and stop the machine in a safe manner, because making decisions based on stale data is dangerous.

ProfiSafe, CIP-Safe, and other safety busses run other Ethernet, have even tighter timing requirements, b/c they involve not machine safety, but human safety.

What this all comes down to is that, if the physical network you run your Profinet over also has to carry file transfers or other heavy-traffic but non-realtime traffic, your network switches must be configured to cleanly handle and prioritize both sets of traffic. Home routers do similar things -- if you have Netflix or video-conferencing running in one room, while someone else is heavily browsing web pages, the "realtime" packets for the "live" services get priority over the "non-live" transfers.

Ideally, your ProfiNet network, and your "file transfer" network, would be physically separate. The KRC4 makes this difficult, since it only has one Ethernet port (KLI) for all this traffic, but normally this doesn't cause the kinds of problems you are describing. However, as the network grows and has more "sprawl," the routing can become very strange, and unpredictable things can happen. I had a case where powering down one cell would cause safety faults on other cells that should have been completely isolated, because the automatically generated routing tables were creating spaghetti -- traffic that should have been isolated inside one cell and never gone further was instead being routed out through multiple other cells and switches, before returning.

I saw at least one major auto manufacturer that failed to understand this, and put all their network traffic (offices, emails, file transfer, Profinet, ProfiSafe, teleconferencing, etc) on the same physical network during a major overhaul, and then spent weeks trying to figure out why all the production equipment kept faulting. In the end, they had to hire some network specialists

Unfortunately, this kind of network know-how can be rather esoteric -- most corporate IT people don't understand the Fieldbus requirements, and most people who really grasp the FieldBus setups don't have a handle on high-end networking, like smart switches, load balancing, etc. This is a gap that is closing over time, but it still exists.

Since many years I change the time to 8 ms on every robot I setup.
Never had any problem.
As already told: interpolation time is 12 ms, so it makes no sense to use a refresh time of 1ms.
I think it is only the lazyness of the system programmer at kuka that makes the default to 1 ms.