How to Choose an Industrial Online Density Meter: A 6-Step Method

Steps

1. Step 1 — Document the medium, including its upsets

   Write down composition, solids content and particle size range, viscosity, temperature range, pH, and whether the process entrains gas. Two entries decide more than the rest: abrasive suspended solids and entrained gas. Abrasion eliminates Coriolis and optical outright and forces a wear-resistant design on anything immersed. Gas degrades transmission-ultrasonic and Coriolis severely. Record the upset range, not just normal operation — startup, washdown, and low-flow periods are when instruments are blamed, and a meter selected only for steady state has no margin for them.

2. Step 2 — Define the accuracy the control loop actually needs

   Ask what decision the number drives. A trend display for operator awareness, a dosing loop, and a custody-transfer measurement have different requirements by an order of magnitude, and buying the third when you need the first is a common and expensive error. Write the requirement as a control tolerance, not as a datasheet figure. Then ask the harder question: over what period, and through what process excursions, does that tolerance have to hold? Short-term repeatability is often what a control loop really needs, and it is a different specification from absolute accuracy.

3. Step 3 — Audit the installation constraints

   Can the pipe be tapped, and is there a shutdown window to do it in? What is the line size? Is there a vertical rising section available, or only horizontal runs? What is the flow velocity range, and does the line ever run part-full or empty? These answers eliminate more candidates than the medium does. Two rules are worth stating outright: a vertical rising section beats a horizontal one because solids stratify at low velocity, and the position immediately downstream of a pump is the worst available spot because turbulence and entrained air both peak there. If tapping is impossible, the feasible set narrows to clamp-on ultrasonic, externally mounted gamma, or a sampling bypass.

4. Step 4 — Check compliance, hazardous area and materials

   Establish the area classification and, critically, which certification scheme your market recognises — ATEX, IECEx, NEPSI, FM and CSA are not interchangeable, and a certificate that is valid where the instrument was built may not be accepted where it will be installed. Confirm this before comparing anything else, because it can eliminate an otherwise ideal candidate. Then confirm wetted materials against the medium at its maximum temperature, not its normal one: 316L that is fine in ambient caustic can stress-corrosion crack in hot concentrated caustic. Add sanitary requirements (surface finish, CIP/SIP, certification) if the application is food or pharmaceutical.

5. Step 5 — Rank the survivors by what will actually go wrong

   You now have a short list where every option is technically feasible. Rank it on the failure modes rather than the specifications: what does each candidate do when the process upsets, and does it tell you it is struggling or fail silently? An instrument that reports a diagnostic — signal quality, a gas index, a temperature fault — lets an operator discount a bad period instead of losing trust in the reading altogether. Then weigh installation cost including any shutdown, spares lead time, and who does service and how far away they are. A two-week spare delivery on a critical loop is a bigger operational cost than a small accuracy difference.

6. Step 6 — Validate on the real medium before committing

   Ask for a calibration on your actual process liquid rather than a reference fluid, and for a trial where the duty is unusual. When the trial runs, compare against laboratory samples rather than against the incumbent instrument, take samples as close to the sensor as the piping allows, and record the time of each — a sampling point far upstream measures transport lag and reports it as instrument error, which is the single most common reason a good instrument fails a trial. Send the completed steps 1–4 to an application engineer with the process data; a selection made from a datasheet alone skips exactly the constraints this method exists to catch.