Also, this is a quench, so I'm only looking at the signal for about 10 seconds. The first second of that is crazy noisy because the water shifts from a bypass valve into the tool, and it takes a bit to settle down.
If I were faced with this measurement task, which sounds like total volume for a short-duration flow event, I would go ahead and try the existing setup with a "field calibration" if practical. By this, I mean setting up a method to capture the total volume of dispensed fluid for a typical quench event, and then correlate actual quench volume with the PLC-computed volume. Then use this relationship to correct the PLC result after the quench is complete.
This would likely give decent results assuming there is nothing terribly wrong with the existing setup (e.g., flowmeter type, flowmeter installation, valve positions, wiring), and other process variables (e.g., fluid pressure, restrictions, nozzles) are reasonably consistent over time.
If taking this approach, capture the fluid using quench periods in the range of typical operation. For example, if it is typically 9 or 10 seconds, capture the fluid for 8-, 9-, 10-, and 11-second quench events. Carefully measure the captured volume and record the PLC-computed quench volume for each test. If possible -- not too expensive or time consuming -- do a couple repeats.
Then use the a spreadsheet to plot the actual volume against computed volume to get a correction equation. MS-Excel is very convenient for this.
Finally program the correction equation in the PLC to correct the flowmeter-measured volume after each quench event. As long as the events are not too far out of the tested time range (and other assumptions hold), the corrected result should be somewhat accurate.
I am assuming quench time is the variable and others stay constant. If time is constant and, say, pressure varies, you would do the same but with different pressure settings.