How to Evaluate Condition Monitoring Solutions as a Plant Manager

The vendor pitch is the same regardless of which plant they are standing in: install sensors, detect failures early, prevent downtime, achieve ROI in months. The pitch does not change when the plant is a tire facility in South Carolina, an appliance plant in Monterrey, or a Tier 1 stamping supplier in the Michigan Auto Alley. The assets change. The failure modes change. The consequences of a missed alert change significantly.

A Banbury mixer gearbox failure in a tire plant shuts down the entire plant immediately, because nothing downstream can run without compounded rubber. A paint shop exhaust fan failure in an appliance plant triggers a finishing line shutdown that starves assembly within hours and creates an EPA compliance issue simultaneously. A stamping press motor failure at a Tier 1 auto parts supplier starts an OEM penalty clock the moment the missed shipment window opens.

Evaluating condition monitoring effectively in discrete manufacturing means evaluating it against your specific assets, your specific failure modes, and the financial consequences of a missed alert in your specific environment. This guide provides the framework.

Start With the Assets, Not the Vendor

Before you contact a single vendor, build your asset priority list. The list drives every subsequent decision: which assets to put sensors on, what failure modes to require coverage for, and which vendor references to request.

In discrete manufacturing, your Tier 1 asset list typically contains:

For tire manufacturing: Banbury mixer motor and gearbox (the plant pacemaker; failure starves every downstream process simultaneously), tread extruder drive, hydraulic power unit pumps serving the curing presses, cooling tower fans and pumps (Banbury overheating from loss of cooling scores batches and damages equipment), and main air compressors.

For appliance manufacturing: Main assembly conveyor drive (every workstation idles on failure), paint booth exhaust fan (finishing line shutdown with EPA compliance implications), stamping press drive motor (primary metal fabrication bottleneck), injection molding machine hydraulic pump motor (plastic component bottleneck), and main air compressors.

For Tier 1 auto parts (stamping): Stamping press main drive motor and transfer system (OEM penalty exposure begins at failure), assembly conveyor drive, cooling tower pump (press overheating from cooling loss stops production), and main air compressors.

For OEM machinery manufacturers: Main assembly conveyor drive, paint shop air handling unit fan, CNC machining center spindles (failure forces part outsourcing at premium cost), and main air compressors.

Once you have this list, every vendor conversation is grounded. The question is not "does your platform detect failures?" It is "does your platform detect gear mesh faults and gear tooth spalling on a Banbury mixer gearbox running at full torque?" or "does your platform detect bearing defects on a stamping press main drive motor under 2,000-horsepower continuous load?" Those are the questions that separate vendors with genuine discrete manufacturing experience from vendors with general industry claims.

The Criteria That Matter in Discrete Manufacturing

Asset-specific failure mode coverage

Vibration monitoring detects the failure modes that matter most in rotating equipment: bearing defects, gear mesh wear, shaft misalignment, and rotor imbalance. For discrete manufacturing, the specific failure modes that are highest-consequence vary by asset type.

On the Banbury mixer gearbox, the failure mode you must catch before it progresses to catastrophic failure is gear tooth pitting and spalling. An early-stage gear tooth defect on a Banbury gearbox, caught and repaired in a changeover window, costs a fraction of what the same defect costs when it progresses to catastrophic failure: a multi-day plant shutdown plus an emergency rebuild at emergency rates. Ask every vendor specifically: how does your AI classify gear mesh fault modes on a high-torque gearbox, and can you show me an alert from a Banbury mixer deployment?

On a stamping press main drive motor, the failure mode is bearing degradation under high-cycle, high-impact load. The load profile is very different from a pump or fan motor: the press cycles at a fixed rate with a high-impact shock load on every cycle. Ask the vendor: how does your platform handle the vibration signal from a high-cycle impact load versus a continuous rotating load? A system that cannot distinguish these will either miss the failure mode or generate false positives on every press cycle.

Diagnostic autonomy

Discrete manufacturing plants do not generally have dedicated vibration analysts on staff. Your maintenance manager and senior technicians understand their equipment intimately, but they are not trained to interpret a Fast Fourier Transform spectrum.

The useful model: when an alert fires, a technician without vibration analysis training receives a notification that identifies the failure mode ("developing bearing defect on the inner race, Motor 7-B on Assembly Line 3"), the asset, the estimated severity, and the recommended action. They confirm the diagnosis with a physical inspection, execute the repair, and log the outcome.

The model that produces unacted alerts: a platform that surfaces vibration waveforms and asks the maintenance team to interpret them. Teams that lack in-house analyst capacity stop investigating these alerts within 30 to 60 days.

During vendor demos, ask to see an actual alert as a technician sees it. If the demo shows a dashboard rather than an alert, push further. The alert format is what determines whether your team will act on it.

Variable-load handling

This is the criterion that discrete manufacturers overlook most often, and the one that determines whether the alert quality is usable after the baseline period.

In a continuous process facility, assets run at roughly constant load. In discrete manufacturing, the load profile changes constantly. An assembly line conveyor runs at full speed during production, at low speed during changeover, and at zero during planned downtime. A stamping press runs at rated speed during production and is idle during tool changes. A Banbury mixer runs at high torque during mixing and at low torque during discharge.

A monitoring platform that does not distinguish between these operating states generates false positives on every normal operational mode change. The team investigates three false positives in week one and stops investigating in week two. The program dies from alert fatigue.

Ask every vendor: how does your platform handle variable-load profiles in discrete manufacturing? What is the mechanism for distinguishing a production vibration signature from a changeover or warmup signature? Ask for customer references who can describe their false-positive rate during changeover cycles.

Integration with maintenance execution

An alert that does not become a work order does not become a repaired asset. The gap between alert generation and work order creation is where most condition monitoring programs lose their effectiveness.

The best implementations integrate the monitoring platform directly with your work order system, either natively with your CMMS or through a defined API. When an alert fires, the work order is created automatically, the parts list is pre-populated based on the failure mode, and the work order is assigned to the maintenance planner's queue. The technician investigates, confirms the diagnosis, and the repair is scheduled for the next available planned window.

If the integration is manual, and a supervisor has to read an email and manually create a work order, you will lose alerts in the gap. Ask every vendor: what does the path from alert to executed work order look like in your customer deployments? Ask references how they handle alerts on Tier 1 assets that come in outside business hours.

Total cost of ownership by monitored point

The hardware price per sensor is the easiest number to compare and one of the least useful ones. The relevant calculation is total annual program cost per monitored critical asset, which includes hardware, software subscription, connectivity infrastructure, onboarding, and ongoing support.

Build your baseline cost per avoided failure event before evaluating any vendor. Pull your last 18 months of unplanned downtime events on your Tier 1 assets. Calculate the average cost of each event: direct production loss, emergency repair premium, and OEM penalty exposure where applicable. The average cost per Tier 1 failure event is the number your program cost should be measured against. A monitoring program that costs $150,000 annually and prevents two Tier 1 failure events per year on assets where each event costs $400,000 is a defensible investment regardless of the hardware price per sensor.

Questions to Ask Every Vendor

These questions distinguish vendors with genuine discrete manufacturing deployment experience from vendors with general manufacturing claims:

1. Can you show me an alert from a Banbury mixer gearbox deployment, or from a stamping press main drive motor? What did the alert say and what did the technician do?
2. How does your platform handle variable-load profiles in discrete manufacturing? What is the mechanism for filtering production versus changeover versus idle states?
3. What is your documented false-positive rate across discrete manufacturing customers in the first 90 days?
4. How does an alert become a work order in your customer implementations?
5. What does support look like when a Tier 1 asset generates an alert at 3 a.m. on a Saturday?
6. Can you provide two references from my specific sub-sector: tire, appliance, auto parts, or OEM machinery? I want to speak with someone who went live 12 to 18 months ago.
7. What failure modes does your platform detect on a gearbox running at high torque versus a standard motor-pump combination?

Where to Start: Asset-Specific Deployment Priority

The plants that see results fastest from condition monitoring are the ones that start narrow and prove the model on their highest-consequence asset before expanding.

Tire manufacturers: Start with the Banbury mixer motor and gearbox. This is the asset where a prevented failure has the highest calculable ROI and the clearest before/after story. A single Banbury gearbox failure avoided, with a documented alert, a confirmed failure mode, and a planned repair in a changeover window, is the internal case study that justifies expanding monitoring to the rest of the plant.

Appliance manufacturers: Start with the main assembly conveyor drive and the paint shop exhaust fan. The conveyor drive is the highest-consequence single point of failure. The paint shop fan is the most common surprise bottleneck: plant managers often underestimate how quickly a paint line failure starves the assembly line.

Tier 1 auto parts (stamping): Start with the stamping press main drive motor and the main air compressors. The press motor failure is your OEM penalty exposure. The compressor failure is your plant-wide shutdown risk. Both should be monitored from day one.

OEM machinery manufacturers: Start with the main assembly conveyor drive and the paint shop AHU fan. For plants with CNC machining centers, add CNC spindle monitoring to the initial deployment: a spindle failure forces part outsourcing at a premium cost that compounds with every day of lead time.

How to Design a Pilot That Produces Useful Data

A pilot is a structured evaluation, not a free trial. Define these three things before the pilot starts:

The target assets. Choose Tier 1 assets that have failure history in your plant: assets that have failed in the last 18 months. An asset that never fails will produce no alerts during the pilot and tell you nothing about the platform's detection capability. If your Banbury mixer gearbox failed 14 months ago, it should be in the pilot set.

Success criteria. Commit to these in writing with the vendor before the pilot begins: confirmed actionable alerts within the pilot window, with a manageable false-positive rate, and at least one failure mode confirmed by technician investigation that your current PM program would not have detected before the pilot period ended.

An internal owner. Assign one person to investigate every alert during the pilot, document what they found, and log the outcome. This person is your pilot owner. Without a dedicated owner, alerts sit unactioned and the evaluation period ends without useful data.

A well-structured pilot runs 60 to 90 days. The first 30 days are baseline establishment; the platform is learning normal signatures for each asset under your production conditions. Expect the first actionable alerts in weeks four to six. Run the pilot through at least one changeover or model transition to test how the platform handles variable load states.

False positive rate, the accountability evaluation criterion: A condition monitoring system that generates frequent false alarms is not a reliability tool, it is an alarm management problem. Every false positive that takes a healthy machine offline for inspection costs production time. Every false positive that gets ignored trains the maintenance team to treat alerts as noise. Ask vendors: what percentage of alerts generated lead to a confirmed fault on physical inspection? At Pirelli, 85% of Tractian alerts were validated as real faults. At Whirlpool, the team documented an 85% insight validation rate. False positive rate is a first-order evaluation criterion for any manufacturing plant running under production pressure.

Pencil whipping prevention, digital accountability: Condition monitoring that produces digital, timestamped alert records solves the manual route accountability problem. Every alert is logged with the asset, the failure mode, the severity grade, the date issued, and the technician response. Unlike a paper route, the data cannot be checked without the inspection being performed. Unlike a spreadsheet, the alert history is visible at the Plant Manager level without waiting for a weekly report. Evaluate whether the platform tracks alert-to-work-order closure rate, that is the accountability metric that proves the monitoring investment is producing maintenance action, not just monitoring data.

Asset lifecycle and CapEx protection: Evaluate whether the platform provides condition trend data at the asset level over the full monitoring period, not just current fault status, but the degradation trajectory over 12–24 months. That historical trend data is the evidence base for condition-based replacement decisions. A Plant Manager who can show a board or plant director that a major asset has 18 more months of service life based on condition trend data is deferring CapEx with evidence. A Plant Manager who replaces on calendar schedule is spending capital based on assumptions. Evaluate the platform's ability to export long-term condition trends as part of any asset replacement business case.