Over the last 40 years of my career, there have been some significant changes in relay technology that have been characterised by certain terms. These terms seem to have been invented/evolved and been adopted over time. Off-hand I am not sure if the terms have ever been "Standard-ised" anywhere (e.g. IEC 60255, IEEE ...???).
So this is my historical recollection and understanding which you might find useful.
Originally there was only ELECTROMECHANICAL which relied on flux as the operating mechanism - even timers and induction disc/cup relays were really just simple motors.
Then came ELECTRONIC, also known as STATIC, with the use of diodes and transistors as individual discrete components on a printed circuit board such as the GEC CTG overcurrent or GEC VTT11 timer.
There were also electromechanical-electronic HYBRID like the GEC YTG distance relay which had a lot of electromechanical components but a small "sealed tin can" with some electronics as the comparator and the cans were filled with a compound for heat stabilisation and environmental protection.
Then in 1981 came DIGITAL with the GEC SHNB Micromho distance relay - the ac waveform was converted into a simple 1 or 0 depending on whether it was the positive or negative half cycle. A couple of years later later there was also the MBCH TF Differential also as a 1/0 representation of the rectified ac waveform being above or below a threshold.
In 1982 came MICROPROCESSOR DIGITAL with the GEC MiDOS range MCGG as the world's first microprocessor protection function device - although it turned the waveform into a "number" effectively representing the r.m.s. value, it simply used a Log table lookup to work out the operating time, i.e. if the number is nnnn, then operating time from the log table is tttt.
Then around the late 80's came NUMERICAL devices with the MiDOS K Series - inherently microprocessor devices using algorithms as numerical calculations to calculate the operation time. As microprocessor based, they also provided the output contact trip matrix logic configuration and also started providing simple communications RS485 options for some sort of setting and/or simple SCADA interface. It is NUMERICAL because the input waveform has been converted into a number representing the magnitude of each individual sample.
In the mid 90's came INTELLIGENT ELECTRONIC DEVICES (IED) where it was definitely microprocessor based but now with "full" Ethernet LAN communications.
And in the early 200's came IEC 61850 based devices as a whole different VENDOR-AGNOSTIC approach to engineering a system and configuring the IEDs to communicate.
Additional clarification SHNB, SHPM, LFZP, EPAC P44x
SHNB Micromho and SHPM Quadramho (mho) are DIGITAL.
The current and polarising waveforms are each converted into a square wave of 01010... purely based on whether waveform is a positive half cycle or a negative half cycle - no magnitude information. The relative sequence of the 0‑to‑1 and 1‑to‑0 transitions then determines if the waveforms represent a fault inside or outside of the zone.
The LFZP Optimho was built on a microprocessor platform (introducing communication capability) but still the same positive/negative 101010 sequence comparison so I would say it is fundamentally a DIGITAL relay or possibly a hybrid MICROPROCESSOR DIGITAL relay None of the three are NUMERICAL.
When you jump to the EPAC and P44x series they are fully NUMERICAL as they use the microprocessor to do algorithm formula calculations on the waveforms that have been converted in an Analog‑to‑Digital converter i.e. each individual sample at say 80 samples per cycle is a binary or hexadecimal NUMBER.
This diagram may assist in understanding the difference between digital and numerical operation ... it shows digital really has no knowledge of instantaneous magnitude whilst numerical has only magnitude information.