Loss-in-weight versus gravimetric control: which one fits your line

First, clear up the terms: the two kinds of "gravimetric" are not the same thing

In the industry, "gravimetric" is used in two places. One is gravimetric feeding — material falls from the hopper into the screw under its own weight — which describes the feeding arrangement. The other is a literal rendering of gravimetric, meaning the whole class of schemes that control on a weight basis, which describes the measurement method. Mix the two meanings and the conversation quickly goes off the rails.

Below we separate them only by the physical location of the measurement:

• Loss-in-weight weight-per-meter unit (loss-in-weight): it weighs the overall weight of the hopper or feeding device, watches the rate at which that weight falls over time, and works back to the extrusion output per unit time. The scale sits upstream of the extruder.
• Gravimetric weight-per-meter controller: it weighs the product itself, or a length cut from it. What is weighed is already-formed cable, on the haul-off and take-up side, downstream of the extruder.

The first difference between loss-in-weight and gravimetric: the two measure at different locations. One is "how much material is about to be used"; the other is "how much has already been used". Response speed and sensitivity to density and to haul-off fluctuation all follow from that.

Three practical differences derived from the difference in location

Response speed and lag

Loss-in-weight measures the flow rate of material entering the die head right now: when the extrusion output changes, the slope of the loss-in-weight curve changes at once, so the system can intervene before the finished dimensions drift. Gravimetric measures cable that has already passed through the cooling trough and completed its run through the haul-off, with cooling and haul-off in between. That time is the lag, and the longer it is, the more the system can only compensate after the fact. Lag is set by the line layout and by line speed, not by the accuracy of the scale — however accurate the scale, it cannot make late information early.

Sensitivity to changes in the bulk density of the raw material

Loss-in-weight tracks the fall in weight, and that quantity is itself accurate. But with volumetric feeding, a change in bulk density changes the weight delivered at the same screw speed: when density falls, the curve shows the extrusion output dropping, and the system raises screw speed until the weight rate returns to target. There the boundary lies too — if density changes too fast, tracking becomes a continuous chase.

Weighing the finished product reads the final weight; bulk density does not enter the reading directly, but the reference value must match the material and be changed with every change of material or batch. Neither method escapes density; it merely appears in a different place — one in the feeding conversion, the other in the setpoint.

Sensitivity to haul-off and take-up fluctuation

Weight per meter is "weight divided by length". Loss-in-weight obtains the extrusion output first and converts it with haul-off speed into weight per meter, so material problems and haul-off problems can be read separately. Finished-product weighing cannot: haul-off slip, tension fluctuation and draw-ratio adjustment all change the weight per unit length, and that is exactly what gets weighed. Adjusting screw speed on that basis will often be wrong in direction — the extrusion output has not changed, the length has. Bringing haul-off speed into the measurement and solving both parameters together softens this, but in essence extrusion fluctuation and haul-off fluctuation are hard to separate, so it is better as a verifier of the result than as the only feedback source.

The two side by side

There is no "better" column. Every row is about the measurement basis, not about accuracy.

Which kind of line suits which one

Leaning towards loss-in-weight: the goal is to keep the extrusion output itself stable. Once it is stable, the haul-off, cooling and take-up that follow have an unchanging input. This suits lines that demand high dimensional consistency and also suffer batch-to-batch variation in incoming material, and situations where volumetric feeding is used and the uncertainty in feeding is to be removed.

Leaning towards finished-product weighing: the goal is to confirm that the material used per unit length meets specification. In material issue, delivery and settlement, which are based on meterage or on weight, the reading can be checked directly against the acceptance basis with no conversion in between, and tracing is simpler. On lines with lower line speed and short cooling and haul-off sections, the lag is relatively manageable.

If the main problem is eccentricity or appearance, neither method helps: weight per meter governs the total amount, not the distribution, and that comes from the die head, the die and concentricity — see where the capability ends in the principles of weight-per-meter control. If the conversion between weight per meter and wall thickness is not sorted out, the setpoint will be wrong whichever measurement method is chosen; settle the relationship between conductor diameter, density and target wall thickness first.

Not an either-or choice: using both at once

Configurations with both on the same line are not uncommon, and the division of labour is usually clear:

• Loss-in-weight gives process-variable feedback at the feed end, holding down short-term fluctuation in extrusion output so that screw speed and haul-off speed have a stable basis for linking.
• Finished-product weight per meter gives reference checking at the take-up end, seeing whether the material that finally emerges really falls inside specification, and giving the feed-end setpoint a direction for correction.

This solves what neither side can solve alone: a stable process does not mean the finished product is right. The feed end can hold the extrusion output very flat, but if the density, shrinkage or wall thickness compensation used in the conversion carries a fixed offset, the finished product is still off as a whole — a "steady but off" condition that process variables alone cannot reveal. Conversely, feedback from finished-product data alone tends to oscillate during transitions because of the lag.

The wall thickness compensation and material usage statistics in the system serve exactly this division of labour: the former handles the fixed offset the conversion introduces, the latter records the weight and meterage on the finished-product basis so that they can be traced back with the process data. Where a line is a multi-section build-up, or several measurement points exist at once, cases can be sorted by operating condition and checked against our solutions. How much compensation to give, and in which direction, has to be found by trial with the actual material.

Four pitfalls in system selection

1. Talking about the two technical routes as one. The proposal says only "gravimetric weight per meter" without stating whether the hopper or the product is measured. Ask one question — "where is the weight signal taken from?" — and most vagueness is exposed on the spot.
2. An accuracy figure with no measurement basis. Accuracy can mean at least four things: the repeatability of the scale, the stability of the control loop, the deviation of the finished-product weight per meter, and whether weight per meter and wall thickness are both within tolerance. The four bases can differ widely, and the qualifier before a specification matters more than the figure.
3. Sensor discussed, actuation not. However high the resolution of the scale, what presses the deviation down is the adjustment of screw speed and haul-off speed. If the actuation cannot keep up, the accuracy stays on the display: the data looks good and the product is out of tolerance anyway. Ask whether an external speed setpoint is available and how large the smallest adjustable step is.
4. Someone else's operating conditions applied to your line. The same scheme shows different lag and stability on lines of different speed, material and cooling section length. Copying specifications without checking these conditions is a common source of recurring transition oscillation.

How to judge which way to choose

Do not ask about schemes first; ask about your own line. Answer these questions and the direction is largely settled:

1. When do deviations mainly occur? Clustered in transitions such as start-up and construction changes, they call for earlier sight of change, and fast feed-end feedback is worth more; spread evenly across the production period, they point to a systematic deviation, so deal with the setpoint and the conversion basis first.
2. Running one construction continuously, does weight per meter drift? Obvious drift points to the stability of the process variable; stable overall but systematically heavy or light points to the setpoint and the compensation.
3. How large is the difference between batches of incoming material? Where density, pellet shape and regrind proportion change often, the feed-end loss-in-weight curve warns earlier that the actual flow rate has changed; with stable incoming material, measurement on the finished-product basis is enough.
4. Are the haul-off and take-up stable? With slipping, tension fluctuation or frequent draw-ratio adjustment, length-side changes mix into the finished-product reading, which then should not be the only feedback source.
5. Can the extruder and the haul-off accept an external speed setpoint, and how large is the smallest adjustable step? If neither is possible, no measurement, however good, can close the loop.

The rest is commissioning: the sampling and repeat-check interval, the compensation amount and the order of linked adjustments must be tuned to the magnitude of your own line's lag. For why a reading swings back and forth, and how to read the swing, see troubleshooting weight-per-meter fluctuation.

Frequently asked questions

What is the core difference between a loss-in-weight weight-per-meter unit and a gravimetric weight-per-meter controller?

Where the weight signal is taken from. Loss-in-weight weighs the overall weight of the hopper or feeding device and reflects how much is being extruded; gravimetric measures the product itself and reflects what has already been extruded. Response speed, sensitivity to density change and sensitivity to haul-off fluctuation all follow from that.

Which of the two is more accurate?

Accuracy without a measurement basis means nothing. The repeatability of the scale and the stability of the control loop are one thing; the deviation of the finished-product weight per meter and whether the finished wall thickness meets specification are another, and they can differ widely. First establish which of these an indicator refers to, and under what conditions it was measured.

We already weigh the finished product — is loss-in-weight still necessary?

It depends on whether the main problem is "steady" or "right". From finished-product data alone, feed-end changes appear only after a lag and transitions tend to need repeated adjustment; from the feed end alone, the process can be very flat, yet a fixed offset in the conversion basis leaves the product off as a whole without showing up. Make up whichever side is the bottleneck first.