How-to
How to program belt feeder ratio blending control on a PLC
Short answer
Program continuous ratio blending by gating the recipe on a 100% ratio sum, computing every feeder setpoint as total rate times its ratio, measuring each belt's flow as belt load times belt speed, and closing one PID loop per feeder onto its VFD speed reference. Verify the blend with accumulated mass per feeder over the same window before releasing product.
On this page
Six belt feeders blending at fixed ratios is one PLC program with three moving parts: a recipe gate that refuses to start until the six ratios sum to exactly 100%, a flow measurement per feeder built as belt load times belt speed, and one closed loop per feeder trimming its VFD speed reference. The operator enters a total rate and a recipe; every feeder setpoint is that total times its ratio. The sections below carry the scaling arithmetic, the setpoint table, the loop structure and the verification test through one worked example: 54 t/h total production, six feeders at 40/30/20/6/3/1 percent.
What a belt weigh feeder measures #
A belt weigh feeder is a gravimetric meter. The weigh bridge under the belt reports the mass of material sitting on the weighed span, the speed sensor reports belt velocity, and the product of the two is instantaneous mass flow; belt scale integrators work exactly this way, deriving flow rate and accumulated weight from those two signals. In units:
Q (t/h) = 3.6 × L (kg/m) × v (m/s)
Kilograms per meter times meters per second gives kilograms per second; multiplying by 3600 seconds per hour and dividing by 1000 kilograms per tonne produces the 3.6 factor. In imperial units the same product is feet per minute times pounds per foot, giving pounds per minute.
Sense speed on a pulley that turns with the belt, not on the motor shaft alone, or motor slip and belt-to-pulley slip become flow errors. Where the speed sensor is a pulse encoder rather than an analog tachometer, the counting happens in a high-speed counter input; configuring a CP1H high-speed counter for an encoder input shows that setup on one controller family.
Draw one boundary early: load times speed is a belt equation. An ingredient dosed by pump or drip tube contributes no belt signal. Measure it with its own flowmeter scaled to the same t/h units, enter its ratio like any other ingredient, and the rest of the program treats it as a sixth feeder.
Program the control in eight steps #
- Enter the recipe as six ratios in percent and sum them in the PLC. If the sum is not 100, latch a recipe-fault alarm and inhibit every feeder enable. Never default a missing entry to zero and never renormalize the others to force the sum.
- Enter the total production rate in t/h. Before allowing a start, compare each computed setpoint with that feeder's commissioned minimum and maximum delivery rates.
- Read each weight transmitter and scale it to kg/m of belt load.
- Read each speed sensor and scale it to m/s.
- Compute measured flow per feeder: Q = 3.6 × L × v.
- Compute all six setpoints: Q_set_i = Q_total × R_i ÷ 100. Recompute only when the recipe or the total rate changes, not every scan.
- Run one PID per feeder: error = setpoint − measured flow, direct acting, output 0-100% as that drive's speed reference.
- Prove the result: run all feeders together and compare each feeder's accumulated mass, and the total, against the entered ratios over the same clock window.
Scale weight and speed into measured flow #
Take feeder 1 of the example. Assume its load-cell transmitter spans 0-40 kg/m of belt load over 4-20 mA, its speed transmitter spans 0-1.0 m/s over 4-20 mA, and the analog input reports 4 mA as raw 0 and 20 mA as raw 27648 — substitute your card's documented counts. The conversion is the ordinary 4-20 mA to engineering-units job; scaling a 4-20 mA analog input to engineering units in CPM2 ladder works it line by line on one PLC family.
Feeder 1 holds 40% of 54 t/h, so its setpoint is 21.6 t/h, which is 21.6 ÷ 3.6 = 6.0 kg/s. At a belt speed of 0.5 m/s the belt should carry 6.0 ÷ 0.5 = 12.0 kg/m. Here is what the PLC sees at that operating point:
- Load input reads raw 8294 (12/40 of span is 8294.4, quantized to whole counts): L = 40 × 8294 ÷ 27648 = 11.998, shown as 12.0 kg/m at one decimal.
- Speed input reads raw 13824: v = 1.0 × 13824 ÷ 27648 = 0.500 m/s.
- Flow: Q = 3.6 × 12.0 × 0.500 = 21.6 t/h.
Measured flow equals setpoint, error is zero, and the loop holds speed. If belt speed fell to raw 11059 (0.400 m/s) with the load unchanged, flow would read 3.6 × 12.0 × 0.400 = 17.3 t/h, and the loop would raise speed until flow came back to 21.6. That is the entire control action.
Validate the recipe and compute the setpoints #
The gate runs before the math. This recipe starts because 40 + 30 + 20 + 6 + 3 + 1 = 100. If an operator edits feeder 3 from 20 to 25 without taking 5 from anywhere, the sum reads 105, the alarm latches and nothing delivers until the recipe is fixed. Six entries or none: a blank field is a fault, not a zero.
At Q_total = 54 t/h the setpoints are:
| Feeder | Ratio | Setpoint |
|---|---|---|
| 1 (main) | 40% | 21.60 t/h |
| 2 | 30% | 16.20 t/h |
| 3 | 20% | 10.80 t/h |
| 4 | 6% | 3.24 t/h |
| 5 | 3% | 1.62 t/h |
| 6 | 1% | 0.54 t/h |
The setpoints sum to 54.00 t/h because the ratios sum to 100.00%.
Two checks sit between this table and a running belt. Capacity: feeder 6 is asked for 0.54 t/h, and on a feeder commissioned for 0-30 t/h that sits near the bottom of its range — compare every setpoint against that feeder's commissioned minimum before start, and give an undersized request a smaller belt or intermittent dosing rather than more loop gain. Reference mode: if total rate should follow what the main feeder actually delivers, set every other feeder to Q_set_i = Q_main × R_i ÷ R_main. With feeder 1 measuring 21.6 t/h, feeder 4 gets 21.6 × 6 ÷ 40 = 3.24 t/h — the same number the table gives when feeder 1 is on target. The follower holds composition when the main feeder wanders; the total-rate scheme holds production rate. Choose one per plant and record it in the recipe.
Close each loop on its drive #
Give the PLC six independent PID instances sampling at the same interval — 1 s is a workable start for belt flow loops, which respond over seconds. Keep the action direct: measured flow below setpoint raises speed. A belt adds dead time, because a speed change reaches the weigh bridge only after material travels from the gate, so start with a modest gain and a few seconds of integral time and tune on the machine; this is the same slow-flow loop shape as S7-300 PID control for a modulating burner motor, if you want a tuning walkthrough.
Clamp each output at 0-100% and freeze the integral while clamped. At startup the belts are empty, measured flow reads zero, and an unwound integral is what slams six drives to maximum on the first automatic scan. Start the drives on a manual or ramped speed and transfer each loop to automatic once its belt shows load.
The speed reference itself is one analog output per drive, 4-20 mA into the drive's reference input; wiring a 1769-OF2 analog output to a PowerFlex 753 input shows that connection terminal by terminal. Network setpoints also work, but six drives polled as slaves on one serial port update slowly — why Modbus RTU drive polling on one serial port is slow gives the arithmetic — so for six loops at a 1 s sample, analog references or a fast fieldbus keep the network out of the loop time.
Keep the conveyor emergency-stop chain and jam protection hardwired so this program never stands between an operator and a stopped belt.
Verify the blend with accumulated totals #
Verification is integrated mass per feeder over a timed window, not a snapshot of percentages. Run the blend steady for ten minutes — one sixth of an hour — while the program totalizes each feeder. Feeder 1 should accumulate 21.6 t/h ÷ 6 = 3.6 t, all six together 54 ÷ 6 = 9.0 t, and 3.6 ÷ 9.0 = 40%, the entered ratio. Feeder 6 gives the same test at small scale: 0.54 ÷ 6 = 0.09 t, which is 90 kg, and 90 ÷ 9000 kg = 1%.
Prove each feeder alone first: force a fixed speed, let the flow calculation settle, then check its computed flow against a physical sample — a cut or diversion weighed on a floor scale. After the single-feeder checks pass, run the combined test. A loop can close tightly around a mis-scaled weight input and hold a wrong flow with perfect confidence; only the accumulated comparison against delivered material catches that.
Questions asked
How does the program change for batch blending instead of continuous?
Keep the same ratio gate and flow measurement, then change the target arithmetic. Enter a batch size in tonnes, multiply it by each ratio for a per-feeder batch target, and integrate each feeder's measured flow. Run each feeder at full rate until its accumulated mass comes within a few percent of target, then drop to a creep speed and stop the drive at target. Composition now depends on the integrators, so verify with accumulated totals per batch.
What do I do with an ingredient that is a liquid?
Treat its pump as a feeder and its flowmeter as its weigh bridge. Scale the flowmeter signal to the same t/h units as the belts, enter its ratio like any other ingredient, and close its own PID around pump speed. Pick the meter for turndown: a liquid at 1% of a 54 t/h blend is asked for 0.54 t/h, about 9 liters per minute of water, which a small meter resolves and a large one does not.
Should the program accept ratios that sum to 99.9 or 100.1?
Only by design, never by accident. If recipes are maintained in whole percents, demand a sum of exactly 100 and hold on anything else, because a 0.1% discrepancy means a stale or mistyped entry. If operators enter two decimals, accept a window such as 99.95 to 100.05 and deliver the values as entered. Either way the alarm text states the rule and shows the sum the program calculated.
Why does the smallest feeder oscillate while the others hold steady?
Turndown. A flow error of a few percent of the total rate is a large fraction of a small feeder's span, so the gain that settles the 40% feeder overdrives the 1% feeder. Lower that loop's gain relative to the others, lengthen its integral time, and check mechanically: at very low belt speeds, material surging from the gate is often the real disturbance. If it still will not settle, the feeder is oversized for its share and should run intermittently.