What Should I Know About Wireless Sensors for Asset Condition Monitoring?

This article by Ed Spence and John Pucillo is Part two of a three-part series discussing wireless Condition Monitoring. Part one was a learning session delivered at the Reliable Plant 2025 conference, with the paper available from Noria Corporation. 

Introduction

Wireless sensors for condition monitoring have the potential to add considerable flexibility to maintenance operations. Judicious deployment of permanently mounted sensors to collect data can save precious time of skilled resources, for example, by monitoring assets that are difficult or unsafe to reach. Analysis of the data can be accomplished from the comfort of your office or even provided remotely as a service. Bad actors or late-stage faults can now be watched more closely by adjusting the measurement time interval with a mouse click. These and other capabilities expand the options available to the maintenance professional, increasing visibility and insight into the condition of critical assets.  

The explosion in wireless products and vendors on the market has been enabled by technologies such as MEMS accelerometers, efficient wireless protocols (such as Bluetooth, with multiple sources of low-cost semiconductor radios), and advanced data processing for diagnostics. Vendors bring a variety of strengths to the market. IoT and software-oriented companies, PdM service providers, industrial equipment manufacturers, sensor manufacturers, instrumentation vendors, and defense-related firms offer a variety of form factors, IT/OT connectivity options, diagnostic tools and sophisticated system features. 

Figure 1. Most new wireless sensors introduced to the market over the last 15 years use embedded, chip-scale MEMS accelerometer technology. 

Wireless vibration sensors exhibit differences with the more familiar accelerometers associated with the handheld data collector, tools long used by a skilled analyst collecting on-demand and route data. Wireless systems overlay additional complexity to the sensor technology, such that it’s worth better understanding how these systems compare. This article will break down the various system components, pointing out features that you ought to be aware of and clarify aspects of the systems that the average user does not need to worry about. 

MEMS Accelerometers—Technology and Specifications

Let’s start with the core sensing technology. Among the key technology developments enabling the rise of wireless systems has unquestionably been the MEMS accelerometer or Microelectro Mechanical System. This chip scale technology attracted early interest of industry observers who recognized the potential for embedded applications with the promise of low power and low cost. Manufactured in scale in semiconductor fabrication facilities, MEMS sensor packages contain highly integrated circuitry that can operate with very low energy consumption, albeit with a range of overall performance. 

Figure 2. This scanning electron microscope image shows the interlocking polysilicon fingers of a MEMS accelerometer proof mass suspended in a near vacuum. Integrated signal processing electronics sense the change in capacitance between both anchored and floating fingers induced by motion along the sensing axis. Courtesy Analog Devices. 

Manual data collection with legacy Piezoelectric accelerometer technology has been practiced for so long that the features and performance of the sensors are taken for granted by experienced users. These same assumptions do not necessarily apply to MEMS accelerometer technology embedded in most wireless sensors today.

New language to describe the behavior and characteristics of the new sensor technology, particularly in the context of a wireless sensor, must be understood. 

For instance, rather than choosing sensors based solely on their sensitivity as traditionally done for route or manual data collection, it is now essential to ensure that the wireless sensor used has an adequate 'Full Scale Range', or the limit of measurable vibration needed to prevent saturation. 

Figure 3. This plot positions available wireless condition monitoring sensors by both Full-Scale Range and 3dB frequency response. Source: The Machine Instrumentation Group. 

As can be seen in Figure 2, all wireless sensors are not created equal. We’ll briefly summarize key specifications, highlighting issues for consideration.

Sensitivity Stability. MEMS brings some attractive characteristics and features to the vibration measurement application. Intrinsic characteristics of MEMS technology include highly stable and repeatable sensitivity over both time and temperature. For the traditional practitioner wondering how often a wireless sensor with an embedded MEMS accelerometer needs to be ‘recalibrated’ once installed, the answer is ‘never’.  

Lower Frequency Response. Unlike legacy piezoelectric technology (PZT) with an IEPE interface (Integrated Electronics Piezo-Electric), MEMS devices will respond to the gravity vector, exhibiting a very low frequency response. However, wireless sensor design tradeoffs may restrict taking full advantage of this feature for very low rotation rate machines. Be sure to check this particular specification if low speed machine monitoring is required. 

Upper (3dB) frequency response. This specification is often taken as a proxy for accelerometer performance when measuring vibration. MEMS based devices designed for vibration measurement cite upper frequency response as low as 1kHz and as high as 23kHz. While many wireless sensors may serve adequately as a ‘check engine’ light, a frequency response that is too low may not satisfy the analyst interested in performing detailed diagnostics, particularly for early detection of emerging bearing fault.

Infamous ‘ski slope’ response. MEMS devices typically exhibit much lower resonant peaks than their PZT counterparts, recovering from saturation in milliseconds. This means that a wireless sensor with embedded MEMS accelerometers is much less likely to suffer from the infamous ‘ski slope’ response of the piezoelectric sensor when subject to shock, such as the impact induced when mounting with a magnet.  

Digital Signal Processing. Wireless sensors are electronic devices which include highly integrated functionality such as digital filters. While intended to be helpful, digital filtering can add an additional layer of uncertainty to the measurement fidelity by obscuring the natural resonant features of the accelerometer response. [1]

Figure 4. The frequency response for two commercially available MEMS accelerometers are compared in this plot. A digital output device from ST Microelectronics IIS3DWB (TOP) exhibits a response shaped by the integrated digital filtering. The response of an analog output device from Analog Devices ADXL1002 (BOTTOM) exhibits resonant peaking at the natural resonant frequency of the accelerometer.  

And of course, the raw analog signal of the MEMS transducer will be ‘digitized’ by a high-resolution ADC somewhere in the signal path, whether embedded in the digital output MEMS device or within the wireless sensor electronics. Design constraints (for example, battery size or amount of embedded memo line of resolution (LOR) – all of which impact overall measurement fidelity and diagnostic flexibility.

Features of Wireless Sensor Systems

Batteries

One of the most vaguely specified and poorly understood aspects of wireless sensor use is battery life. Battery life is a complex function of environmental variables such as temperature as well as user configured operation such as measurement frequency (interval).

Many vendors cite battery life measured in years, albeit under conditions that are not often clearly articulated. Battery life can be estimated by measuring actual energy consumption of the sensor under various conditions such as temperature and transmission distance. Some products offer a battery charge indicator much like your mobile, but estimation of battery State-of-Charge (SOC) is a difficult and complex exercise. [2] More than one client has approached us for help when installed sensors demonstrated battery life inconsistent with the specified performance, increasing support costs in the field. Battery life prediction is an imprecise science and vendor specifications in general are an area that could use some improvement across the industry. 

Figure 5. Primary cell battery duration comparison for the Saft LS 14250 Lithium Thionyl Chloride (TOP) and Panasonic CR2477 Lithium Manganese Oxide (BOTTOM), both used in commercially available wireless vibration sensors. Trying to use the voltage output alone as a gauge for available energy capacity would be misleading. 

Mechanics

Best practices regarding sensor mounting on the asset or component have not changed. Sensors mounted with stud or adhesive will provide repeatable measurements with more fidelity, particularly important for sensors specifying a higher frequency response. Sensors citing a lower frequency response of say, 1kHz, will not be as sensitive to potential shifts in resonant response due to magnet mounting, enabling faster deployment. Magnets should still have sufficient pull strength and contact on the measurement surface to ensure solid mechanical coupling to the asset under all conditions. 

One aspect of mechanical response that has changed is changes to the frequency response of the wireless sensor due to enclosure design. Vertical height, mechanical design, material choices, MEMS component mounting on the Printed Circuit Board (PCB) and the addition of battery mass theoretically influence an overall enclosure frequency response. Any added mechanical resonances may be more observable on products with higher frequency response, whereas the impact might be negligible for the many lower frequency response products on the market. Without making measurements ourselves, we’re relying on the reputation and mechanical design experience of the vendor to achieve the specified measurement fidelity necessary for the application. Frequency response plots common to accelerometer component datasheets are not nearly as common for wireless sensors.  

 

Figure 6. This frequency response plot was measured for a client prototype submitted to us for testing. Despite using a MEMS device with 3db bandwidth specified greater than 5kHz, resonances introduced by the mechanical design are observable at lower frequencies. This product was consequently not introduced to the market in the US. Source: The Machine Instrumentation Group.  

Vendor Profiles

Wireless vendors can be classified into two broad categories. Many market entrants include technology or software-oriented enterprises with competencies in IoT, and all this implies - RF connectivity, embedded firmware, as well as web-based dashboards and often “AI” or data driven analysis tools. These companies have necessarily had to learn something about machine mechanics and vibration measurement by hiring experienced vibration analysts. 

In the other camp are industrial instrumentation or PdM service providers who were already part of the condition monitoring eco-system and adding IoT systems to their existing tool kits. Many of the providers already have positions inside the end user facilities, with boots on the ground taking route-based measurements with familiar instrumentation tools and trained by the same eco-system. 

These developments have added several new features and specifications specific to wireless networks to the condition monitoring instrumentation. We will briefly review those that ought to be of interest to the maintenance and reliability user.

Network Topology and IT / OT interface. Most wireless systems on the market today are deployed with ‘star’ network configurations, which include centrally located gateways to collect transmitted data from multiple sensors. Gateways then reformat the data for passing along a wired connection to a PLC or the facility LAN. Many gateways on the market today provide both WIFI and Ethernet connections, as well as a cellular connection that can be provisioned with a SIM card (cellular subscription), generally to simplify deployment (avoid wrangling with IT).   

Transmitted data is transferred via LAN to a dashboard hosted either on a commercial web hosting service (such as AWS) or an internal server, both accessible with appropriate login credentials via laptop or PC. Many vendor software platforms have provisioned for connection to other platforms in common use by maintenance such as CMMS and the plant historian, generally with cloud-to-cloud software (REST API). 

Vendors with maintenance or process control experience may also provision the gateway with connections to the PLC through traditional OT protocols such as Modbus. Beyond that, much of the networking information provided by radio-oriented vendors regarding the RF transmission is of limited use to the end user. Do you really need to be concerned about whether the sensor transmits to the gateway with a peak transmission power of 8dBm? Only so far as the choice of protocol, along with many other system design choices, impacts the cost of deployment and quality of network coverage within the facility. For example, the number of sensors supported by a gateway may, all else being equal, have an impact on the number of gateways needed for deployment, potentially impacting the capital cost of the deployed system. Vendors with proprietary protocols may differentiate their products by supporting a high number of sensors per gateway or enable very short transmission times that claim to support longer battery life. 

Web-based dashboards. System configuration of sensor locations, asset hierarchy, etc. is often performed initially with a mobile app during installation but can also be managed from the system configuration page of the dashboard.  Data dashboards provided by the vendor can enable drilling down through a facility hierarchy to assets with color coded operational status, beneath which there is available trended condition indicator plots, frequency spectrum and other diagnostic tools for the analyst. Diagnostics, with generation of evaluation reports or even work orders may be performed remotely by the vendor as part of an offered service or alternatively by internal resources. Several vendors boast data driven or ‘AI’ diagnostic capabilities to scale the number of sensors supported but are quick to reassure the customer that human experts touch every analysis report. Data driven approaches leverage both the massive amount of data collected by the vendor as well as the expertise of skilled human analysts helping to train the machine learning algorithms. While fully automated diagnostics may have achieved tentative confidence for specific assets and applications, automation of the diagnostic decisions for a broader base of assets still require human supervision, at least while we all gain confidence in data driven alerting and diagnostics. 

Diagnostics and AI. It seems more and more systems on the market today are ‘smart’. The term ‘AI’ has been so often used as to have little meaning for most end users. Should we consider a system leveraging a database or library of industrial component specifications such as bearings ‘artificially intelligent’? Classical physics-based analysis of signal processing, such as fault frequency detection, can also deliver diagnostic information if enough is known about the asset being monitored.

Diagnostic accuracy of any system takes time and experience to fully assess. Vendors often highlight faults captured with little explanation of the process used or the training involved. Diagnostic approaches may use machine learning techniques during a training period to recognize faults as well as set alarm thresholds. A lab test comparison of three comparable ‘smart’ sensors sponsored by a client yielded mixed results – after running the devices on an industrial component test platform for the specified training period, one product performed as advertised, one misclassified an induced mechanical fault, and one device would not train due to a software bug. Caveat emptor. 

Track Record and Trends

Wireless solutions began emerging at least 15 years ago, with a few early entrants growing into the market leaders of today while other early products have not survived. Currently we believe that nearly 1 million sensors have been deployed, led by those same leaders but now augmented by a 2nd tier of companies with a variety of backgrounds. Hard market data is difficult to come by, but investors and vendor C-suite executives privately disclose that 80-90% of the market is still unpenetrated. New entrants seeking to penetrate North American facilities continue to appear from Brazil, Turkey, Israel, India, China and Europe.

Given the complexity of these systems, a description of the system features or assurances from marketing collateral alone is not a reliable gauge of network performance or cost of deployment in any given facility. The actual cost of deployment depends also on intangibles such as the high density of metal structures, moving vehicles and other interferers typical of an industrial environment. And of course, the sheer variety of assets that could be monitored suggests that it’s unlikely there is a one-size-fits-all solution, a most vendors offer only vibration and temperature. [3]

The cost of ownership will depend also on the flexibility of the vendor service models offered, including any additional services offered. Many new entrants to the market have focused on remote vibration analysis as-a-Service, whereas traditional PdM service providers may blend their wireless deployment costs into billable hours for other (route-based) services. The optimum business model depends heavily on the maturity of the buyer’s PdM program and should consider the competency levels of the internal maintenance and reliability teams. 

Conservative adopters will be interested in the financial soundness, breadth of resources, geographic reach, experience and PdM pedigree of the vendor or service provider. Testimonies from current customers are invaluable, and some leading vendors make references readily available, regularly hosting summits full of customer testimonials.

A thorough study of vendor collateral comparing specifications and features, supplemented by vendor interviews by knowledgeable individuals will narrow your options, enhancing the chances of finding a good fit, but the only sure way to validate spec sheets or sales claims is to deploy and evaluate in your environment. Given the time, energy and cost of testing a system, many first-time adopters solicit recommendations from trusted sources before engaging. 

 

In Part three of this series, we consider how the reliability and maintenance team may effectively approach the daunting task of curating and engaging wireless condition monitoring vendors.