Design and Preliminary Experimental Characterization of a LowCost ESP32-Based System for Electric Motor Condition Monitoring

Small and medium-sized manufacturing enterprises without digital instrumentation detect incipient faults in electric motors by sight and sound, which leaves no record of the vibration and temperature changes that precede a breakdown. This study designed and preliminarily characterized a low-cost prototype, based on an ESP32 microcontroller, that measures vibration and temperature and classifies the motor state against approximate thresholds derived from the International Organization for Standardization (ISO) 10816-1 class limits. It also reports a preliminary proof of concept of a tiny machine learning (TinyML) anomaly detector intended to run on the microcontroller. The research was quantitative, applied and cross-sectional, in two stages. In stage 1, the prototype, with an MPU6050 accelerometer sampled at 20 Hz and a DS18B20 sensor, measured five workshop motors once each. The three motors with a known rated speed fell 71% to 85% below the approximate warning threshold (0.05 to 0.08 m/s² root mean square, RMS), a non-validated classification; a portable drill recorded 1.36 m/s² but could not be classified for lack of a rated speed. In stage 2, the sensor noise floor was characterized at 500 Hz, and a first capture of a running lathe motor gave an RMS 4.6 times that floor, with most energy between 65 and 130 Hz rather than at the rotation frequency. Because faults could not be induced, an INT8 autoencoder of about 3.9 kilobytes was trained on the 155 windows of that single session and tested with synthetic anomalies, reaching a sensitivity of 0.69 and a specificity of 0.95 with a chronological split; the threshold rule detected none. These optimistic figures were not measured on the microcontroller. Low-cost acquisition is feasible, but threshold conversion requires spectral velocity analysis before it can be considered equivalent to the standard. A bench calibration found gain errors below 0.6%, and the detector runs on the microcontroller in about 49 µs per window using 0.01% of the processor time

Authors

Institutions

Publication Details

Journal
International Journal of Research and Innovation in Social Science
Published
2026-11-14
DOI
https://doi.org/10.47772/ijriss.2026.102500007
Primary Topic
Machine Fault Diagnosis Techniques
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Design and Preliminary Experimental Characterization of a LowCost ESP32-Based System for Electric Motor Condition Monitoring

Yael Omar Acosta Rangel, Diego Ortega Pérez, Noel Montiel Elizalde, Gael Rangel Castañón et al.
International Journal of Research and Innovation in Social Science
Machine Fault Diagnosis Techniques
article

Design and Preliminary Experimental Characterization of a LowCost ESP32-Based System for Electric Motor Condition Monitoring

Yael Omar Acosta Rangel, Diego Ortega Pérez, Noel Montiel Elizalde, Gael Rangel Castañón, Irving Yahir Barrera Nochebuena
article en

Abstract

Small and medium-sized manufacturing enterprises without digital instrumentation detect incipient faults in electric motors by sight and sound, which leaves no record of the vibration and temperature changes that precede a breakdown. This study designed and preliminarily characterized a low-cost prototype, based on an ESP32 microcontroller, that measures vibration and temperature and classifies the motor state against approximate thresholds derived from the International Organization for Standardization (ISO) 10816-1 class limits. It also reports a preliminary proof of concept of a tiny machine learning (TinyML) anomaly detector intended to run on the microcontroller. The research was quantitative, applied and cross-sectional, in two stages. In stage 1, the prototype, with an MPU6050 accelerometer sampled at 20 Hz and a DS18B20 sensor, measured five workshop motors once each. The three motors with a known rated speed fell 71% to 85% below the approximate warning threshold (0.05 to 0.08 m/s² root mean square, RMS), a non-validated classification; a portable drill recorded 1.36 m/s² but could not be classified for lack of a rated speed. In stage 2, the sensor noise floor was characterized at 500 Hz, and a first capture of a running lathe motor gave an RMS 4.6 times that floor, with most energy between 65 and 130 Hz rather than at the rotation frequency. Because faults could not be induced, an INT8 autoencoder of about 3.9 kilobytes was trained on the 155 windows of that single session and tested with synthetic anomalies, reaching a sensitivity of 0.69 and a specificity of 0.95 with a chronological split; the threshold rule detected none. These optimistic figures were not measured on the microcontroller. Low-cost acquisition is feasible, but threshold conversion requires spectral velocity analysis before it can be considered equivalent to the standard. A bench calibration found gain errors below 0.6%, and the detector runs on the microcontroller in about 49 µs per window using 0.01% of the processor time

International Journal of Research and Innovation in Social ScienceVol. 10(25)
Tecnológico Nacional de México, Campus Pachuca (MX)
Machine Fault Diagnosis Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon