Glosario de monitoreo de máquinas y IA industrial: 20 definiciones esenciales

Understanding the language of complete machine monitoring guide and industrial AI is the foundation for making informed decisions about technology, communicating with vendors, and building internal alignment around monitoring programs. This glossary defines the 20 most important terms in the industrial AI and machine monitoring space, written for manufacturing professionals — not data scientists.

Which Industrial AI and Machine Monitoring Terms Do Manufacturers Need to Know?

OEE guide">Overall Equipment Effectiveness (OEE)

The manufacturing industry’s gold-standard metric for production efficiency, combining availability, performance, and quality into a single percentage.

OEE is calculated as Availability × Performance × Quality. Availability measures uptime vs. planned production time. Performance compares actual output rate to the theoretical maximum. Quality measures first-pass yield as a percentage of total output.

World-class OEE is considered 85% or above. Most manufacturers who begin measuring OEE automatically (rather than manually) discover their true OEE is 10–20 points below their previous estimates, because manual measurement misses micro-stops and speed losses.

Why it matters: OEE is the primary KPI that machine monitoring software improves. Every percentage point of OEE improvement on a production line translates directly to recovered revenue without adding capital or headcount.

Unplanned Downtime

Any machine stoppage during scheduled production time that was not anticipated or planned.

Unplanned downtime is distinguished from planned downtime (scheduled maintenance, changeovers, setup) by its unscheduled nature. It is caused by equipment failures, material problems, process faults, and human factors.

Unplanned downtime is the primary target of machine monitoring and predictive maintenance strategy programs because it is largely preventable with the right data. Industry benchmarks suggest that 30–50% of unplanned downtime events show detectable precursor signals in sensor data 24–96 hours before the stoppage occurs.

Why it matters: Unplanned downtime is the most expensive and most preventable form of production loss. Machine monitoring converts it from a surprise event to a predictable, manageable maintenance action.

Predictive Maintenance (PdM)

A maintenance strategy that uses real-time condition data to predict when equipment needs service — before failure occurs but not before it’s necessary.

Predictive maintenance differs from preventive maintenance (time-based service regardless of condition) by using actual machine condition data to determine when maintenance is needed. This eliminates both unnecessary maintenance (performing service on equipment that doesn’t need it) and unexpected failures (not performing service until after breakdown).

Predictive maintenance requires machine monitoring infrastructure to collect the condition data that models analyze. The monitoring data — vibration spectra, thermal profiles, current draw patterns — contains the early warning signals that enable prediction.

Why it matters: Predictive maintenance programs typically reduce unplanned downtime by 30–50% and maintenance costs by 10–25%. They are the highest-ROI application of machine monitoring data.

Mean Time Between Failures (MTBF)

The average time a machine or component operates between failures under normal conditions.

MTBF is the primary reliability metric for production equipment. A machine with an MTBF of 2,000 hours is expected to fail approximately once every 2,000 operating hours under normal conditions. MTBF is used to plan maintenance intervals, predict spare parts needs, and benchmark equipment reliability over time.

MTBF data from machine monitoring builds automatically as the monitoring system accumulates failure history. Comparing MTBF by machine type, age, and operating conditions reveals which equipment is underperforming relative to design specifications or fleet peers.

Why it matters: Improving MTBF is the operational outcome of a successful predictive maintenance program. Higher MTBF means fewer failures per operating hour, which means less downtime, lower maintenance cost, and more predictable production.

Mean Time to Repair (MTTR)

The average time required to restore a failed machine to operational status, measured from the moment of failure to the moment production resumes.

MTTR measures maintenance team effectiveness — how quickly can the team diagnose, source parts, and repair a failed machine. MTTR differences between facilities with similar equipment often reveal gaps in diagnostic capability, spare parts availability, or technical skills.

Machine monitoring reduces MTTR in two ways: real-time failure notification (reducing the time between failure and maintenance team awareness) and condition context at the time of failure (reducing diagnostic time by providing the sensor history leading up to the failure).

Why it matters: Reducing MTTR reduces the production impact of failures that do occur. A facility that reduces average MTTR from 2.5 hours to 45 minutes on a 10-machine floor with one failure per machine per month recovers 175 hours of production per year.

Condition-Based Maintenance (CBM)

A maintenance strategy that triggers service actions based on the actual measured condition of equipment, rather than on a fixed time or usage schedule.

Condition-based maintenance is the practical implementation of the predictive maintenance philosophy. Instead of servicing every machine on a fixed schedule (every 500 hours, every 3 months), CBM triggers service when sensor data indicates the machine’s condition has degraded to a threshold that warrants intervention.

CBM requires monitoring infrastructure — sensors and analytics capable of measuring and trending the condition parameters relevant to each failure mode. Machine monitoring platforms like SensFlo implement CBM by continuously trending vibration, temperature, and other condition indicators and generating alerts when trends exceed defined thresholds or deviate from established baselines.

Why it matters: CBM eliminates both over-maintenance (servicing machines that don’t need it) and under-protection (waiting for failure on machines that do). The combination typically reduces total maintenance cost by 15–25%.

Vibration Analysis

The measurement and interpretation of mechanical vibration data to assess the condition of rotating machinery components.

All rotating machines — motors, pumps, fans, compressors, spindles — vibrate. The frequency and amplitude of that vibration contains information about the condition of the machine’s internal components. Bearings, gears, imbalanced rotors, and misaligned shafts each produce characteristic vibration signatures that can be detected and trended over time.

Bearing defect frequencies — BPFI (ball pass frequency inner race), BPFO (ball pass frequency outer race), BSF (ball spin frequency), and FTF (fundamental train frequency) — are mathematical functions of the bearing geometry and rotational speed. AI-based vibration analysis detects rising amplitude at these specific frequencies, indicating bearing degradation.

Why it matters: Vibration analysis is the most powerful and most widely applicable predictive maintenance technique. Virtually every machine with rotating components benefits from vibration monitoring, and bearing failures — detectable through vibration months before catastrophic failure — are the most common preventable failure mode in manufacturing.

IoT (Industrial Internet of Things)

The network of physical sensors, devices, and machines connected to the internet for the purpose of collecting, transmitting, and analyzing operational data.

The Industrial IoT (IIoT) refers specifically to the application of IoT technology in industrial and manufacturing settings. IIoT devices include vibration sensors, temperature sensors, current transducers, flow meters, and environmental sensors that attach to production equipment and transmit data wirelessly to cloud platforms.

IIoT has democratized machine monitoring by reducing the cost and complexity of sensor hardware from enterprise-level capital projects to accessible subscription-based services. A sensor that cost $2,000 and required specialist installation in 2015 now costs $200 and installs in 60 seconds.

Why it matters: IIoT is the enabling technology for machine monitoring, predictive maintenance, digital twins in manufacturing, and AI-driven factory management. Without IIoT sensors, manufacturing AI has no real-time data to work with.

OPC-UA and industrial protocols-opc-unified-architecture">OPC-UA (OPC Unified Architecture)

The modern standard communication protocol for secure, platform-independent industrial machine data exchange.

OPC-UA is a machine-to-machine communication protocol developed by the OPC Foundation that allows industrial machines to securely expose their operational data to external systems. Unlike older protocols (Modbus, PROFIBUS), OPC-UA includes built-in security (authentication, encryption), supports complex data structures, and is platform-independent.

Most CNC machines manufactured after 2015, modern injection molding presses, industrial robots, and other Industry 4.0-compatible equipment support OPC-UA. A machine monitoring platform that can read a machine’s OPC-UA server can access the controller’s full data: spindle speed, axis positions, part counts, program states, alarm histories.

Why it matters: OPC-UA integration enables machine monitoring platforms to access the rich operational data that lives inside machine controllers, complementing the physical condition data from external IoT sensors.

Digital Twin

A continuously updated virtual model of a physical asset that uses real-time data to mirror the asset’s current state and predict its future behavior.

A digital twin is not a 3D CAD model (which is static) or a sensor dashboard (which shows current data without a model). A true digital twin uses live sensor data, control system telemetry, and operational history to maintain a dynamic model of a machine or system — one that can be used to simulate, predict, and optimize behavior.

Machine monitoring is the data foundation of a digital twin. Without continuous, high-quality sensor and telemetry data, a digital twin cannot maintain an accurate current-state model. SensFlo provides the real-time condition data layer that keeps digital twins synchronized with physical machines.

Why it matters: Digital twins enable capabilities that data dashboards alone cannot: remaining useful life prediction, virtual commissioning, remote expert diagnosis, and simulation-based process optimization.

Six Big Losses

The six categories of production loss that OEE is designed to measure and quantify, as defined in Total Productive Maintenance (TPM) methodology.

The Six Big Losses are: (1) Equipment Failure (unplanned breakdowns), (2) Setup and Adjustments (changeover and setup time), (3) Idling and Minor Stops (small stops under 5–10 minutes), (4) Reduced Speed (operating below maximum rated speed), (5) Process Defects (scrap and rework during stable production), and (6) Reduced Yield (startup scrap and quality losses during process changes).

Each of the Six Big Losses corresponds to one of OEE’s three components: Losses 1–2 are Availability losses, Losses 3–4 are Performance losses, and Losses 5–6 are Quality losses. Machine monitoring software quantifies all six categories automatically, enabling targeted improvement actions.

Why it matters: The Six Big Losses framework is the analytical foundation for OEE-driven improvement programs. Understanding which loss category is driving an OEE gap points directly to the right intervention.

AI-driven anomaly detection

The automated identification of patterns in sensor data that deviate significantly from established normal behavior, indicating potential equipment problems.

AI-based anomaly detection is the core technology behind predictive maintenance alerts. Instead of comparing current readings to fixed thresholds, anomaly detection algorithms learn the normal operating pattern of each machine — its typical vibration signature, temperature profile, and current draw at different loads and speeds. When data deviates from this learned baseline in ways that are statistically significant, an anomaly alert is generated.

Unsupervised anomaly detection (used by SensFlo) requires no labeled training data or manual threshold configuration. The AI establishes baselines automatically from normal operating data and begins flagging anomalies without human configuration. This makes it practical for any manufacturing environment, regardless of whether historical failure data is available.

Why it matters: Anomaly detection catches the gradual degradation patterns that threshold monitoring misses. The most expensive manufacturing failures — spindle bearing failures, hydraulic pump failures, gearbox failures — develop gradually over weeks or months before catastrophic failure. Anomaly detection catches them in the early stages.

CMMS (Computerized Maintenance Management System)

Software for planning, tracking, and documenting maintenance activities, work orders, spare parts, and equipment history.

A CMMS is the maintenance team’s operational system of record. It tracks work orders (both planned and unplanned), maintenance history by asset, spare parts inventory, technician assignments, and maintenance cost. Common CMMS platforms include Fiix, Limble, Hippo, Infor EAM, and SAP PM.

Machine monitoring platforms like SensFlo generate the data that drives CMMS work orders. When SensFlo’s AI detects a developing bearing failure, the alert can automatically create a predictive maintenance work order in the CMMS with machine context, sensor data, and recommended action — eliminating manual data entry and ensuring the maintenance team acts on AI-generated insights.

Why it matters: Integration between machine monitoring platforms and CMMS systems closes the loop between detecting problems and fixing them, with full documentation of the condition data that triggered each maintenance action.

Machine Utilization Rate

The percentage of scheduled production time that a machine is actively running and producing output.

Machine utilization rate is distinct from OEE: utilization measures whether the machine is running, while OEE measures whether it is running productively (at the right speed, producing good parts). A machine that is available and running but making scrap has high utilization and low OEE.

Utilization rate is most useful as a capacity planning metric: are machines underutilized (available capacity exists) or fully utilized (bottleneck risk is high)? Machine monitoring provides accurate utilization data by detecting run state in real time — something that production scheduling systems typically estimate rather than measure.

Why it matters: Understanding utilization rates across a machine fleet reveals hidden capacity and identifies bottleneck machines before they become delivery constraints.

Takt Time

The rate at which products must be produced to meet customer demand — calculated as available production time divided by customer demand volume.

Takt time is the manufacturing industry’s fundamental pacing metric. If a customer demands 480 units per day and 480 minutes are available for production, takt time is 1 minute per unit. Every machine and process in the production system must operate at or faster than takt time to meet demand.

Machine monitoring supports takt time adherence by detecting cycle time drift in real time. When a machine’s actual cycle time rises above takt time, an alert signals that the line is falling behind before the production shortfall becomes a delivery problem.

Why it matters: Takt time monitoring is particularly valuable in assembly lines and flow production environments where one slow station affects the entire line’s throughput.

Edge Computing (Industrial Edge)

The processing of sensor and machine data at or near the source of data generation (the factory floor) rather than transmitting all raw data to a central cloud.

In industrial IoT deployments, edge computing uses local hardware (an edge gateway, industrial PC, or embedded controller) to process raw sensor data before it leaves the facility. This reduces network bandwidth requirements, enables local alerting that works without cloud connectivity, and reduces latency for time-sensitive monitoring applications.

SensFlo’s edge gateway performs local data processing and buffering — aggregating sensor data, running initial anomaly detection, and storing data locally during network interruptions. This ensures no data is lost during connectivity outages and enables the 90-second alert response time that overnight monitoring requires.

Why it matters: Edge computing is the architectural feature that makes machine monitoring reliable in manufacturing environments where network connectivity is inconsistent. Local processing ensures monitoring continues to function and data is preserved regardless of cloud connectivity status.

Spindle Utilization

The percentage of scheduled machine time that a CNC machining center’s spindle is actively cutting (rotating and engaged in the workpiece), as distinct from setup, idle, and air-cutting time.

Spindle utilization is the CNC machining industry’s primary productivity metric — analogous to OEE’s performance component but specific to cutting time. A machining center with a spindle utilization of 45% is producing chips for less than half its scheduled time. World-class CNC operations target 70–85% spindle utilization.

Improving spindle utilization requires reducing non-cutting time: setup duration, tool change time, air-cutting time, and idle time between jobs. Machine monitoring tracks spindle utilization automatically by detecting spindle motor current — cutting produces a distinct current signature compared to air cutting or idle.

Why it matters: Spindle utilization improvement is the primary productivity lever for CNC job shops. A job shop that improves spindle utilization from 45% to 65% on a 10-machine floor at $150/hour machine rate adds $1.5M in annual capacity without capital investment.

Preventive Maintenance (PM)

Scheduled maintenance performed at regular time or usage intervals to prevent equipment failure, regardless of the machine’s actual condition.

El mantenimiento preventivo es la estrategia de mantenimiento más practicada en la fabricación: el servicio en un horario (cada 500 horas, cada 3 meses, cada 10,000 ciclos). El PM es mucho mejor que el mantenimiento puramente reactivo (arreglarlo cuando se rompe), pero tiene dos limitaciones fundamentales: el exceso de mantenimiento (servicio de máquinas que aún no lo necesitan) y la subprotección (máquinas que fallan entre intervalos PM porque su condición se degradó más rápido de lo previsto).

El mantenimiento preventivo y el mantenimiento predictivo son complementarios, no mutuamente excluyentes. La mayoría de los fabricantes utilizan PM para equipos de baja criticidad y enfoques predictivos de reserva (basados en condiciones) para activos de alto valor y alta criticidad donde la falla es más costosa.

Por qué es importante: Los programas PM proporcionan una línea de base de disciplina de mantenimiento planificado. El monitoreo predictivo se basa en esta línea de base al identificar activos de alta criticidad que justifican enfoques basados en condiciones, y al identificar intervalos de PM que son demasiado largos (activos que frecuentemente fallan antes de su fecha de PM) o demasiado cortos (activos que están constantemente en buenas condiciones a la hora de la PM, lo que indica un exceso de mantenimiento).

MTConnect

Un protocolo de comunicación estándar abierto diseñado específicamente para permitir que las máquinas herramienta CNC compartan datos operativos con sistemas externos en un formato estandarizado e interoperable.

MTConnect es administrado por la Association for Manufacturing Technology (AMT) y proporciona un flujo de datos XML estandarizado que expone los datos de la máquina CNC, incluida la velocidad del husillo, las posiciones de los ejes, las tasas de alimentación, los recuentos de piezas, la información de la herramienta, el estado de ejecución del programa y los códigos de alarma. A diferencia de los protocolos patentados (Fanuc FOCAS, Mazak MAZATROL), los datos de MTConnect son independientes de la marca de la máquina: la misma integración funciona en Mazak, Okuma, Mori Seiki, Hardinge y otras marcas.

MTConnect está cada vez más disponible en máquinas CNC de los principales fabricantes, y los adaptadores de retrofit están disponibles para controles Fanuc y Siemens más antiguos. Es el estándar de integración preferido para el monitoreo de máquinas CNC cuando está disponible.

Por qué es importante: MTConnect permite que las plataformas de monitoreo de máquinas accedan a los ricos datos del controlador que hacen que el monitoreo CNC sea profundamente perspicaz: cargas reales del husillo, recuentos de piezas reales, seguimiento de rendimiento a nivel de programa e historial de códigos de alarma, datos que los sensores externos por sí solos no pueden proporcionar.

Retorno de Asset Intelligence (ROAI)

La métrica patentada de SensFlo para cuantificar el retorno financiero generado por los datos de monitoreo de máquinas, ampliando el concepto tradicional ROA (Retorno sobre Activos) para incluir el valor de inteligencia de las máquinas conectadas.

El Retorno de Activos Tradicional (ROA) mide el rendimiento financiero relativo al valor de los activos. ROAI amplía esto al cuantificar la cantidad de retorno adicional que se genera cuando los activos se conectan al monitoreo en tiempo real, capturando el valor de producción recuperado a partir de un menor downtime, los costos de capital evitados por el mantenimiento predictivo y las ganancias de eficiencia de la mejora de OEE.

SensFlo's Calculadora ROAI permite a los fabricantes ingresar el recuento de máquinas específico, el valor de la máquina, la frecuencia de downtime y el valor de producción por hora para calcular el ROAI esperado a partir de una implementación de monitoreo. Los clientes reportan un ROAI de 3 a 10 veces en el primer año.

Por qué es importante: ROAI es el marco para comunicar el ROI del monitoreo de máquinas a los tomadores de decisiones financieras. Traduce los beneficios operacionales del monitoreo (menos desglosamientos, mejor OEE, menor costo de mantenimiento) al lenguaje financiero que requieren las decisiones de inversión de capital.

Preguntas frecuentes

P: ¿Qué es OEE en términos simples?

OEE (Total Equipment Effectiveness) es un porcentaje que le indica cuánto tiempo de producción planificado es verdaderamente productivo. Un OEE del 100% significa que la máquina funcionó todo el tiempo, a toda velocidad, haciendo solo buenas piezas. La mayoría de los fabricantes puntajes entre 40 y 60% cuando comienzan a medir por primera vez, lo que significa que existe un potencial de mejora significativo.

P: ¿Cuál es la diferencia entre el mantenimiento predictivo y el prescriptivo?

El mantenimiento predictivo le indica cuándo es probable que una máquina falle en función de los datos del sensor. El mantenimiento prescriptivo va más allá: le indica exactamente qué acción tomar y cuándo. El asistente de IA FloE de SensFlo ofrece recomendaciones prescriptivas en un inglés sencillo, no solo umbrales de alerta.

P: ¿Qué significa el mantenimiento basado en condiciones (CBM)?

El mantenimiento basado en la condición significa dar servicio al equipo solo cuando los datos de monitoreo indican que es necesario, no en un calendario fijo. CBM reduce tanto las tareas de mantenimiento innecesarias como las fallas no planificadas mediante el uso de datos reales del estado de la máquina para impulsar las decisiones.

P: ¿Qué es IIoT vs IoT?

IoT (Internet de las cosas) se refiere ampliamente a los dispositivos conectados. IIoT (Internet Industrial de las Cosas) es el subconjunto aplicado a equipos industriales — máquinas de fabricación, sensores y sistemas que transmiten datos operativos para análisis y automatización. El monitoreo de máquinas es una aplicación básica de IIoT.

P: ¿Qué es un gemelo digital en la fabricación?

Un gemelo digital es un modelo virtual de una máquina o proceso físico, actualizado continuamente con datos reales del sensor. Los gemelos digitales permiten la simulación de escenarios de falla, la optimización de los parámetros operativos y el monitoreo remoto del estado de la máquina sin requerir acceso físico.

Lectura relacionada

¿Listo para comenzar? Solicite una demostración gratuita— la mayoría de los fabricantes están monitoreando sus primeras máquinas dentro de una semana. Utilice el Calculadora ROAI para proyectar su devolución, o explorar precios para encontrar el nivel adecuado para su operación. Más información sobre nuestras soluciones, FLoe AI, y casos de éxito de clientes.