There are many ways to add intelligence to industrial systems, including edge and cloud artificial intelligence (AI) matched to sensors with analog and digital components. Given the diversity of AI approaches, sensor designers must consider multiple competing requirements, such as latency for decision-making, network usage, power consumption/battery life, and AI models suitable for machines. This article highlights the application of intelligent AI-powered wireless motor monitoring sensors and the related solutions introduced by ADI.
Motor health monitoring using wireless industrial sensors
Condition-based monitoring (CbM) for robots and rotating machinery (such as turbines, fans, pumps, and motors) records real-time data related to machine health and performance, enabling targeted predictive maintenance and optimized control. Targeted predictive maintenance early in a machine's lifecycle reduces production downtime risks, thereby improving reliability, significantly cutting costs, and enhancing efficiency on the factory floor. Industrial machine CbM can utilize a range of sensor data, including electrical measurements, vibration, temperature, oil quality, acoustics, magnetic, as well as process measurements like flow and pressure. However, vibration measurement is by far the most common, as it provides the most reliable indicators of mechanical issues such as imbalance and bearing failures.
Wireless industrial sensors currently on the market typically operate at very low duty cycles. Users set the sensor's sleep duration, after which the sensor wakes up to measure temperature and vibration, then transmits the data via radio signals back to the user's data aggregator. Commercial sensors often claim a battery life of five years, based on collecting data once every 24 hours or multiple times within that period.
In most cases, the sensor spends over 90% of its time in sleep mode. Take ADI's Voyager4 sensor as an example, it operates similarly but uses edge AI anomaly detection (with the MAX78000 AI microcontroller) to limit radio usage. When the sensor wakes up and measures data, it transmits the data back to the user only if the microcontroller detects an anomaly, triggering machine diagnostics and maintenance while extending motor life. By using AI at the edge, battery life can be extended by at least 50%.
Voyager4 is a wireless condition monitoring platform developed by ADI, designed to help developers quickly deploy and test wireless solutions for machines or test setups. Voyager4 and other motor health monitoring solutions are widely used in robotics and rotating machinery such as turbines, fans, pumps, and motors.
How the voyager4 sensor system works
The Voyager4 sensor is paired with the ADXL382, a triaxial 8 kHz digital microelectromechanical system (MEMS) for collecting vibration data. First, the raw vibration data is transmitted to the MAX32666, a low-power Bluetooth® (BLE) processor. The data can be sent to the user via BLE radio or USB, and this raw vibration data is used to train an edge AI algorithm using the MAX78000 tools.
Using the MAX78000 tools, the AI model is synthesized into C code. The edge AI algorithm is sent to the Voyager4 sensor via a BLE over-the-air (OTA) updates and stored in memory using the MAX78000 processor with edge AI hardware accelerator. After the initial training phase of Voyager4, the ADXL382 MEMS data can be transmitted along the path. The MAX78000 edge AI algorithm predicts machine faults or normal operation based on the collected vibration data. If the vibration data is healthy, the MAX32666 radio is not used, and the MEMS returns to sleep mode. However, if abnormal vibration is predicted, a vibration anomaly alert is sent to the user via BLE.
In Voyager4's hardware system, the ADXL382 is a low-noise-density, low-power, three-axis MEMS accelerometer with selectable measurement ranges. The device supports ±15 g, ±30 g, and ±60 g ranges and a wide measurement bandwidth of 8 kHz. The ADG1634 is a single-pole double-throw (SPDT) CMOS switch used to route raw MEMS vibration data to either the MAX32666 BLE radio or the MAX78000 AI microcontroller. The BLE microcontroller controls the SPDT switch. Several other peripherals are wired to the MAX32666, including the MAX17262 fuel gauge for monitoring battery current and the ultra-low-power ADXL367 MEMS accelerometer. The ADXL367 is used to wake the BLE radio from deep sleep mode during a high-vibration shock event, consuming only 180 nA in motion-activated wake-up mode. The BLE microcontroller can transmit ADXL382 MEMS raw data to the host via BLE or USB through the FTDI FT234XD-R.
The Voyager4 sensor uses the MAX20335 power management integrated circuit (PMIC), which features two ultra-low quiescent current buck regulators and three ultra-low quiescent current low-dropout (LDO) linear regulators. Each LDO and buck regulator output voltage can be individually enabled or disabled, and each output voltage value can be programmed via I²C with the default preconfigured. The BLE processor is used to enable or disable individual PMIC power outputs for different Voyager4 operating modes.
In training mode, the BLE microcontroller must first advertise its presence on the BLE network, then establish a BLE connection with the network manager. The Voyager4 then transmits ADXL382 MEMS raw data over the BLE network to train the AI algorithm on the user's PC. Afterward, the Voyager4 sensor returns to deep sleep mode. In normal (AI) mode, BLE radio advertising, connection, and streaming functions are disabled by default. The MAX78000 periodically wakes up to run an AI inference. If no anomaly is detected, the Voyager4 returns to deep sleep mode.
ADI's Voyager4 evaluation kit (EV-CBM-VOYAGER4-1Z) includes several components (LEDs, pull-up resistors) for customer evaluation. These components generate a deep sleep current of 0.3 mW on the LDO1OUT voltage rail. The average power consumption of the Voyager4 evaluation kit is calculated based on the time intervals between events in deep sleep, training, and normal/AI modes.
The following sections further detail the functional characteristics of these related components.
AI microcontrollers enable neural networks to run at the edge of the IoT with ultra-low power
The MAX78000 is an AI microcontroller featuring an ultra-low-power convolutional neural network accelerator. This new AI microcontroller enables neural networks to operate at the edge of the IoT with ultra-low power consumption, combining energy-efficient AI processing with Maxim's proven ultra-low-power microcontrollers. With this hardware-based convolutional neural network (CNN) accelerator, even battery-powered applications can perform AI inference while consuming only microjoules of power. The MAX78000 is an advanced system-on-chip integrating an Arm® Cortex®-M4 with FPU CPU , enabling efficient system control through an ultra-low-power deep neural network accelerator. The device is packaged in a 81-pin CTBGA (8 mm x 8 mm, 0.8 mm pitch).
The MAX32666 is a low-power ARM Cortex-M4 FPU-based microcontroller with Bluetooth 5, designed for wearable applications. This next-generation UB MCUs are engineered to meet the complex demands of battery-powered and wireless-connected devices. It pairs with larger memory compared to similar products, adopting a scalable memory architecture. The device uses wearable-grade power technology for long-lasting operation and durability, capable of withstanding high levels of cyberattacks. It is available in 109-bump WLP (0.35 mm pitch) and 121-bump CTBGA (0.65 mm pitch) packages.
The ADXL382 is a low-noise, low-power, wide-bandwidth, three-axis MEMS accelerometer with selectable measurement ranges, supporting ±15 g, ±30 g, and ±60 g. The ADXL382 offers industry-leading noise performance, enabling precision applications with minimal calibration. Its low noise and low power consumption allow for accurate measurement of audio signals or heart sounds even in high-vibration environments. The ADXL382's multifunction pin names can be referenced either by their functions related to the serial peripheral interface (SPI) or I²C interface or by their audio functions (pulse-density modulation (PDM), I²S, or time-division multiplexing (TDM)). The ADXL382 comes in a 14-terminal, 2.9 mm x 2.8 mm x 0.87 mm LGA package.
A complete solution for wireless asset health monitoring using edge AI
Voyager4 leverages edge AI for wireless asset health monitoring. Voyager4 uses triaxial digital output MEMS sensors, including the ADXL382 and ADXL367. The design also incorporates the MAX32666 BLE and MAX78000 AI microcontrollers. Flexible and PCB area efficient PMIC power devices are added as load switches to enhance the energy efficiency of wireless sensors. Each Voyager4 kit includes a BLE 5.3 adapter with an antenna. Voyager4 uses BLE and is compatible with any PC equipped with a Bluetooth radio, but for optimal performance and range, it is recommended to use the adapter when communicating with Voyager4.
The ADG1633/ADG1634 is a 4.5 Ω RON, three-/four-channel, single-pole double-throw (SPDT), ±5 V/+12 V/+5 V/+3.3 V switch. The ADG1633 and ADG1634 are monolithic industrial CMOS (iCMOS®) analog switches, each integrating three/four independently selectable single-pole double-throw switches. All channels feature break-before-make switching to prevent momentary shorting during channel switching. The ADG1633 (LFCSP and TSSOP packages) and ADG1634 (LFCSP package only) provide an EN input to enable or disable the device. The iCMOS structure ensures extremely low power consumption, making these devices ideal for portable, battery-powered instruments.
The ADXL367 is a nanopower, three-axis, ±2 g/±4 g/±8 g digital output MEMS accelerometer. At a 100 Hz output data rate, it consumes only 0.89 µA, and in motion-triggered wake-up mode, it consumes just 180 nA. Unlike accelerometers that use power duty cycling to achieve low power, the ADXL367 does not alias input signals through undersampling but samples the sensor's full bandwidth at all data rates. The ADXL367 is packaged in a 2.2 mm x 2.3 mm x 0.87 mm form factor.
The MAX17262 is a 5.2 µA, ModelGauge m5 EZ single-cell battery fuel gauge with built-in current sensing. It is the industry's lowest IQ fuel gauge, featuring an integrated current-sense resistor and ModelGauge m5 EZ algorithm, eliminating the need for battery characterization. The MAX17262 monitors single-cell battery pack with integrated current sensing, capable of detecting pulse currents up to 3.1 A. The IC is optimized for battery capacity measurement ranging from 100 mAh to 6 Ah. The MAX17262 comes in a tiny, lead-free, 0.4 mm pitch, 1.5 mm x 1.5 mm, 9-pin wafer-level package (WLP).
The MAX20335 is a compact PMIC with ultra-low IQ voltage regulators and a battery charger for lithium-ion systems, featuring an optimized power management solution for 24/7 monitoring systems such as wearables and IoT. The MAX20335 battery charging management solution is ideal for low-power wearable applications. The device includes a linear battery charger with a smart power selector and several power-optimized peripherals. The MAX20335 is packaged in a 36-bump, 0.4 mm pitch, 2.72 mm x 2.47 mm wafer-level package (WLP).
Conclusion
Microcontrollers with built-in AI hardware accelerators provide wireless sensor nodes with better decision-making capabilities and longer battery life. By using AI at the edge, battery life can be extended by at least 50%. Vibration sensors incorporating modal analysis accelerate sensor development cycles and ensure high-quality vibration data capture from monitored assets. ADI's Voyager4 wireless condition monitoring platform, paired with related component solutions, will be your best ally in adding intelligence to industrial systems.