I think that your problem is that the AWS NLB can't do a health check using UDP, so you will need to do the health check using TCP or HTTP/HTTPS.
You can do that adding a TCP input that will be used only for this healthcheck, than in your target group you configure the health check to use this port.
Or you can try this experimental feature that exposes the metrics http endpoint, and use this endpoint as the target for the health check.