Question 1

What sensors should be used to detect a bearing failure?

Accepted Answer

For detecting bearing failures, three types of sensors work together. Piezoelectric accelerometers (IEPE type, ICP standard) measure vibrations in the acceleration range from 0.5 Hz to 10 kHz with a sensitivity of 100 mV/g—they detect the characteristic BPFO (Bearing Pass Frequency Outer), BPFI, BSF, and FTF characteristic of each bearing model. PT100 or NTC temperature sensors mounted on the bearing housing detect abnormal heating (typical threshold: +15 degrees Celsius relative to operating temperature). Motor current analysis (MCSA—Motor Current Signature Analysis) via a 4–20 mA current clamp enables the detection of bearing faults in the current spectrum without mechanical contact. Eziwan natively integrates IEPE sensors via a BNC-compatible vibration acquisition module, and 4–20 mA analog sensors directly on the gateway’s terminals.

Question 2

What is the difference between preventive and predictive maintenance?

Accepted Answer

Scheduled preventive maintenance is triggered by time (every X operating hours or every X months), regardless of the equipment’s actual condition. It replaces parts that are still in good condition and necessitates planned downtime that is sometimes unnecessary. Predictive maintenance, on the other hand, is triggered by the actual condition of the equipment: action is taken when sensor data indicates actual deterioration. In practice, a bearing with a theoretical service life of 20,000 hours may fail at 8,000 hours under severe conditions, or last 30,000 hours under optimal conditions. Preventive maintenance falls short in both cases. IIoT-based predictive maintenance detects rising temperatures or changes in the vibration spectrum 2 to 6 weeks before a failure occurs, allowing maintenance to be scheduled at the optimal time.

Question 3

How long does it take to see a return on investment from predictive maintenance?

Accepted Answer

In our customer deployments, the return on investment is typically achieved within 6 to 18 months. The first significant savings generally occur when the first unplanned shutdown is avoided: for a food production line, a 4-hour shutdown represents 60,000 euros in lost production plus the costs of emergency repairs. The Eziwan predictive maintenance subscription for this same site ranges from 400 to 800 euros per month, depending on the number of pieces of equipment being monitored. Mathematically, avoiding just one shutdown per year justifies the investment. Added to this are savings on parts: no longer replacing bearings that are still in good condition represents 15 to 25% savings on the spare parts budget. Our industrial customers are seeing a 50 to 65% reduction in MTTR (Mean Time To Repair) thanks to early warnings that allow them to prepare for service calls.

Question 4

Does the solution work with our existing equipment?

Accepted Answer

Yes, Eziwan's philosophy is total backward compatibility. For devices with 4–20 mA or 0–10 V analog outputs (most industrial sensors from the 1990s to the present), the connection is made directly to the gateway’s terminals without any changes to the installation. For PLCs and drives using Modbus RTU over RS-485 or Modbus TCP over Ethernet, integration is read-only and requires no changes to the PLC program. For equipment without any digital or analog outputs (old motors, legacy pumps), retrofit sensors are added: a non-contact temperature sensor on the housing, a surface-mounted accelerometer, and a current clamp on the power supply. The installation of these additional sensors is non-intrusive and does not require a production shutdown.

Question 5

Do you need a data scientist to analyze data?

Accepted Answer

No. Eziwan is designed for maintenance managers and technicians, not for data scientists. Establishing the baseline (reference values under normal operating conditions) is automatic: the system learns the normal behavior of each piece of equipment during the first 7 to 14 days and automatically sets adaptive alert thresholds. Deviation (rate-of-change) alerts can be configured in just a few clicks via the web interface: “alert if the temperature rises by more than 2 degrees Celsius per hour.” The dashboards are pre-built with standard maintenance KPIs: MTBF, MTTR, operational availability, and the number of corrective vs. preventive interventions. For advanced users who wish to develop custom ML models, the REST API and the ability to export raw data to Python or industrial analytics platforms (Azure IoT Analytics, AWS IoT SiteWise) are available on the Enterprise plan.

Question 6

How can we integrate alerts with our CMMS?

Accepted Answer

Eziwan offers three methods for CMMS integration, depending on your level of sophistication. The simplest method is the HTTP webhook: when an alert is triggered, Eziwan sends a JSON-formatted HTTP POST request to your CMMS endpoint—this automatically creates a service request or work order. The second method uses Eziwan’s bidirectional REST API: your CMMS periodically queries the Eziwan API to retrieve new alerts and integrate them into its workflow. The third method is native integration with major CMMS systems: pre-configured connectors are available for Mobility Work (a French cloud-based CMMS), SAP PM via SAP PI/PO, and IBM Maximo via the Maximo Application Framework REST API. For on-premises CMMS systems (Infor EAM, CARL Source, Dimo Maint), integration is handled via Eziwan’s ETL module, which generates importable CSV/XML files at regular intervals.

Question 7

Predictive Maintenance vs. Preventive Maintenance: What ROI Can We Expect?

Accepted Answer

A financial comparison between scheduled preventive maintenance and IIoT-based predictive maintenance shows that predictive maintenance is more cost-effective across all industrial sectors with critical equipment. The formula for calculating ROI is as follows: ROI = (Savings Generated - Cost of the Solution) / Cost of the Solution × 100. The savings generated consist of three components. First component: reduction in unplanned downtime. An unplanned shutdown on an automotive or food production line costs between 10,000 and 100,000 euros in lost gross margin per hour (source: Aberdeen Group and ARC Advisory studies). Avoiding two 4-hour stoppages per year on a line costing 15,000 euros per hour represents 120,000 euros in annual savings. Second benefit: reduced spare parts costs. Preventive maintenance results in the replacement of 20 to 30% of parts that are still in good condition. With an annual parts budget of 80,000 euros, predictive maintenance saves 16,000 to 24,000 euros by replacing only parts that are actually worn out. Third area: reduction in technician time. Advance preparation for service calls (ordering parts, resource planning) reduces MTTR (Mean Time To Repair) by 40 to 60%. Concrete example: For a plastics factory with 20 monitored motors, the cost of the Eziwan solution is 8,400 euros per year (subscription + gateway). Estimated savings in the first year: 1 avoided shutdown (30,000 euros) + savings on parts (12,000 euros) = 42,000 euros. ROI = (42,000 - 8,400) / 8,400 = 400%. Industry studies (Deloitte, McKinsey) confirm an average ROI of 10x over 3 years for predictive industrial maintenance.

Question 8

What types of sensors should be used to monitor motors, pumps, and compressors?

Accepted Answer

The choice of sensors depends on the type of equipment being monitored and the available budget. For electric motors (the most common category), the optimal combination is: a triaxial MEMS accelerometer (vibration along 3 axes, measured in g, typical range 0–8 g, 0–3 kHz) to detect bearing faults and imbalances, a PT100 or PT1000 bearing temperature sensor (resistance temperature detector, range -50 to +200 degrees Celsius, accuracy 0.1 degrees Celsius) to monitor overheating, and optionally a 4–20 mA current clamp (MCSA) to analyze the current’s spectral signature and detect broken rotor bars or eccentricities. For centrifugal pumps, add a 4–20 mA differential pressure sensor at the inlet and outlet (process pressure gauge such as WIKA or Endress+Hauser Cerabar): a deviation of the operating point from the HQ curve indicates impeller clogging or incipient cavitation. For reciprocating compressors, oil pressure (4–20 mA transmitter on the lubrication circuit) and cylinder temperature (Type K thermocouple on the cylinder head) are the critical parameters. For transformers and electrical panels, non-contact infrared sensors (FLIR, Optris) via USB or Ethernet interface, or PT100 probes integrated into the panels, can detect overheating in branch circuits upstream of the circuit breaker. All these sensors connect to the Eziwan gateway via direct 4–20 mA analog inputs or via a multi-input Modbus RTU module.

Question 9

How long does it take to get the first reliable predictions?

Accepted Answer

The process of obtaining reliable predictions consists of three successive phases. Phase 1—Baseline and Training (4 to 8 weeks): The system collects nominal operating data to establish a baseline for each piece of equipment. During this period, Eziwan learns normal operating patterns: temperature ranges under load and at idle, typical vibration levels at different rotational speeds, and start-up and shutdown cycles. It is important that this phase cover representative operating conditions (various load conditions, seasonal ambient temperatures if possible). Phase 2 — Calibration of alert thresholds (2 to 4 weeks): After the baseline phase, the system proposes initial adaptive thresholds. The maintenance team adjusts them based on the criticality of the equipment and risk tolerance. A safety motor in an ATEX zone will have more conservative thresholds than a tertiary ventilation motor. Any initial false positives are corrected during this phase. Phase 3 — First Actionable Predictions (2 to 4 months after launch): Drift detection models (trend analysis over 30–90 days) begin to generate alerts predicting abnormal trends several weeks in advance. It is during this phase that the operational value becomes tangible. Accelerating factor: If you have historical data on past failures (dates, nature of faults, observed values), importing this data into Eziwan significantly speeds up calibration by providing labeled examples of actual failures.

Industrial Predictive Maintenance: Anticipate breakdowns before they halt your production

Corrective, Preventive, or Predictive Maintenance: Which One Should You Choose?

Corrective Maintenance

Preventive Maintenance

IIoT Predictive Maintenance

From Vibration to Alert: How IIoT Predictive Maintenance Works

Sensor Acquisition

Secure Transportation

Analysis and Detection

Alerts and CMMS Integration

Industrial equipment compatible with Eziwan predictive maintenance

Electric motors

Pumps and Compressors

Conveyors and Mixers

HV/EHV Transformers

Generator Sets

Industrial HVAC

How much does an hour of unplanned downtime cost you?

Process Industries

Automobile

Agroalimentaire

Manufacturing Industries

Compatibility of Sensors and Industrial Protocols

Supported Protocols

Validated Sensor Brands

Standards and Norms

Everything You Need to Know About IIoT Predictive Maintenance

Get started with your predictive maintenance pilot in 5 business days