How to Maximize Recovery Rates with Steinert Sensor-Based Sorters: A 5-Step Field Checklist

Who This Checklist Is For

If you're running a Steinert magnetic separator or a sensor-based sorter like the Steinert KSS or Steinert ISS, and you're chasing that last 2-3% recovery without blowing your budget on re-tweaks and downtime, this is for you. I've been in quality management for mineral processing and recycling lines for a while now—reviewing specs, signing off on first articles, and rejecting setups that don't meet our internal standards. This checklist is based on what I've seen work (and fail) across roughly 200 unique installations over the last 5 years.

We'll go through 5 concrete steps. One of them—step 4—is something most operators get wrong. I'll call it out.

Step 1: Establish a Baseline Feed Characterization

You can't optimize what you haven't measured. Before touching any parameters on your Steinert unit, you need a solid feed profile.

What to do:

• Run a full particle size distribution (PSD) analysis on your current feed. Use sieves or an automated analyzer.
• Document the moisture content—especially important for sensor-based sorters. (I learned this the hard way in 2022: a feed at 8% moisture vs. 4% completely changed the detection profile.)
• Identify the target material's shape distribution (flaky vs. round vs. irregular). This affects belt loading and sensor detection probability.

Checkpoint: Do you have at least 3 representative samples from different shifts? If not, your baseline is skewed. A single grab sample won't cut it.

People think the machine's fancy sensors will just magically sort everything. But if your feed's variability is wild—say, the PSD swings from 10mm to 80mm within an hour—the sorter's algorithms can't keep up. The assumption is that the sorter adapts instantly. The reality is it needs a consistent feed picture for its training model to hold. (Should mention: this is where some operators skip the step and blame the equipment later.)

Step 2: Configure Material Liberation and Presentation

This is about how the material hits the belt or chute. On a Steinert KSS or similar, presentation is way more important than most manuals let on. Seriously, we've seen a 15% recovery difference just from tweaking the slide angle and chute width.

What to do:

• Check the belt speed first. Rule of thumb: for granular materials (5-20mm), start at 2-3 m/s. For coarser (20-80mm), go lower, 1-2 m/s. But verify this with a high-speed camera—your eyes lie at those speeds.
• Optimize the material monolayer. A double layer on the sensor belt kills detection. Adjust the vibratory feeder amplitude until you get a consistent single layer of particles. This is a no-brainer, but I've seen it missed on every third site visit (ugh).
• Eliminate cross-belt distribution issues. Use a width sensor or manual adjuster to ensure material isn't biased to one lane.

Checkpoint: Is the belt loading uniform within ±5% across the width? If you have a 'shadow lane' where nothing passes, you're leaving money on the table.

Step 3: Tune Sensor Sensitivity and Thresholds

This is where the black magic lives. Steinert sorters use XRF, NIR, CMOS cameras, or magnetic induction sensors. Each has a sweet spot.

What to do:

• Create a calibration sample set from your own feed: collect 20-30 good items and 20-30 bad items (reject materials). Run them one by one—watch the software's heatmap or signal strength.
• Set the reject threshold conservatively at first: you want to miss some bad stuff rather than throw out good stuff. Then iterate. (This feels counterintuitive to operators who want zero impurities—it's a trade-off.)
• Enable statistical classification if your Steinert unit supports it. In our Q1 2024 quality audit, we found that using the default threshold without statistical averaging increased false rejects by 12%.

Checkpoint: Have you run a blind test with your calibration set? Our standard: 95%+ recovery with <2% contamination in the product stream. If you're hitting 98% recovery but with 5% contamination, the threshold is too aggressive.

People think 'zero contamination' is the goal. Actually, the question is: what's your downstream process willing to accept? For a pre-concentrator, 5% contamination might be fine. For final product, it's a deal-breaker. The causation goes the other way—don't set the machine to a spec you don't actually need.

Step 4 (The One Most People Miss): Validate Ejector Timing and Air Pressure

Here's the step that gets overlooked. You've tuned the sensors, the belt speed is right, the material looks beautiful on the belt. But the air ejectors or reject system is off by milliseconds. You're wasting compressed air and misdirecting particles.

What to do:

• Measure the latency between detection signal and ejector activation. On a Steinert unit, this should be <20ms. Anything above 30ms and you're losing good material.
• Check the air knife pressure at the valve. Not at the compressor—at the valve. Pressure drop across the line can be 0.5-1 bar, and that changes the blow force significantly. (Circa 2023, we found a 0.6 bar drop at one site; the line was undersized by 1/4 inch.)
• Inspect the ejector nozzle alignment weekly. A misaligned nozzle by even 2mm can shift the trajectory of the particle by 5cm at the product splitter. We rejected a first delivery in 2024 because the nozzle alignment was off by 4mm—it cost us a $22,000 redo and delayed the launch.

Checkpoint: Are you on a regular schedule for ejector calibration? Monthly is minimum for high-throughput lines. Trust me, this is better than tuning the sensors again.

Step 5: Document the 'Region-Specific' Feed Variance

What works in your plant in Arizona may not work in your facility in Germany. This isn't about the machine—it's about the ore or scrap source.

What to do:

• Keep a feed variance log per source or supplier. Track particle shape index, moisture, and impurity types. This is the dataset you'll use when you get a new batch and want to pre-set your Steinert parameters.
• Run a fast recalibration (Steps 1-4) every time you change feed sources. In our experience, a new supplier's material can shift the recovery curve by 5-8% if you don't adjust.

Checkpoint: Do you have a template for this log? If not, create one with a few key fields: date, source id, PSD, moisture, target recovery, actual recovery. It's a game-changer for consistency.

Common Pitfalls to Avoid

• Over-relying on the 'Auto-Tune' feature. It's a starting point, not a final setup. In 2022, we had a vendor claim their auto-tune would hit 97% recovery. It hit 91% on our feed. Always verify with a real-world test.
• Ignoring compressed air quality. Moisture or oil in the air can clog ejectors mid-run. Install a good dryer and filter—it's cheap insurance.
• Running sorters without a belt cleaner. Build-up on the belt affects sensor readings. (Ugh, I see this way too often. Clean the belt at the start of every shift.)

This checklist was accurate as of early 2025. Mineral processing technology evolves fast—verify current Steinert software versions and component specs before locking in your SOP. (The Steinert Practice Center offers on-site calibration support, which I'd recommend for your first few runs. They helped us cut our setup time by 40%.)

Bottom line: follow these steps, track your recovery data, and you'll get that extra margin without expensive redesigns. Seriously—give it a try on your next shift.