High-Fidelity Vibration Acquisition Platform for Condition Monitoring

【Introduction】This article explains how recent advances in MEMS technology are bringing accelerometers to the forefront to compete with piezoelectric sensors in condition monitoring applications; it also discusses how to use the new development platforms that make this possible.

Introduction to Condition Monitoring (CbM) and Predictive Maintenance (PdM)

Condition monitoring (CbM) involves the use of sensors to measure the current state of health to monitor machines or assets. Predictive maintenance (PdM) requires a combination of techniques such as CbM, machine learning and analytics to predict future asset maintenance cycles or possible failures. Global device health monitoring is expected to grow significantly, making it imperative to be aware of and understand key trends. More and more CbM companies are adopting PdM to improve the differentiation advantage of their products. Maintenance and facility managers now have new options for CbM, such as wireless units, as well as lower-cost high-performance units. While the infrastructure for most CbM systems remains the same, we can now integrate new MEMS technologies directly into systems that previously used mostly piezoelectric sensors, or were not monitored due to cost barriers.

Condition Monitoring – Engineering Challenges and Design Decisions

In a typical CbM signal chain design, there are many different engineering specifications and techniques that need to be considered, all of which are constantly improving and increasing in complexity. There are now various types of clients who may have expertise in a certain area, such as algorithm development (software only) or hardware design (hardware only), but not always proficient in both.

For developers who want to focus on algorithm development, they require a data repository that can accurately predict asset failures and downtime. They don’t want to design hardware, or fix data integrity failures; they want to use really high fidelity data. Likewise, for hardware engineers looking to increase system reliability or reduce costs, they need a solution that can easily connect to existing infrastructure, allowing them to benchmark existing solutions. They need access to data in a readable format that is easy to use and export so that they don’t waste time evaluating performance.

Many of the system-level challenges can be solved with a platform approach—from the sensor all the way to algorithmic development—enabling all customer types.

Many system-level challenges can be addressed with a platform approach (from sensors to algorithm development), supporting all types of customers.

What is CN0549? How does it help extend the life of the device?

CN0549 CbM Development Platform

The CN0549 Condition Monitoring Platform is a high-performance, off-the-shelf hardware and software solution for streaming high-fidelity vibration data from assets into an algorithm/machine learning development environment. The platform provides hardware experts with a tested and proven system solution that provides highly accurate data acquisition, reliable mechanical coupling to assets, and high-performance broadband vibration sensors. All hardware design files are also provided to help you easily integrate into your designed product. Also attractive to software experts, CN0549 outlines the condition monitoring signal chain hardware challenges, allowing software teams and data experts to directly start developing machine learning algorithms. Important features and benefits include:

● Easy installation into assets while maintaining mechanically coupled signal integrity

● Wide bandwidth MEMS accelerometer sensor with IEPE data output format

● IEPE, high-fidelity data acquisition (DAQ) solution with analog input bandwidth from DC to 54 kHz

● Embedded gateway captures and stores raw data for local or networked processing

● Real-time Display of frequency domain data using ADI’s IIO oscilloscope application

● Stream sensor data directly to popular data analysis tools such as Python and MATLAB®

The CbM development platform is mainly composed of four different components (shown in Figure 1), which we will introduce one by one, and then introduce the entire combined solution.

High-Fidelity Vibration Acquisition Platform for Condition Monitoring

Figure 1. Components that make up the CbM development platform

Highly accurate, high-fidelity data capture and processing

With wider bandwidth and lower sensor noise, faults such as bearing problems, cavitation and gear meshing can be detected earlier. It is very important that the data acquisition electronics ensure high fidelity of the measured vibration data; otherwise, important fault information may be lost. Ensuring the fidelity of vibration data allows us to spot trends faster and provide predictive maintenance recommendations with great confidence, reducing unnecessary wear and tear on mechanical components, which in turn extends the life of your assets.

A Cost-Effective Approach to Condition Monitoring of Less Important Assets

Piezoelectric accelerometers are the highest performance vibration sensors used on the most critical assets, where performance is more important than cost. The high cost of piezoelectric sensors has historically prevented condition monitoring of less critical assets. MEMS vibration sensors are now comparable to piezoelectric sensors in terms of noise, bandwidth, and g-range, allowing maintenance and equipment managers to gain greater insight into less critical assets that were previously subject to troubleshooting or passive maintenance plan. This is mainly due to the high performance and low cost of MEMS. We can now use a cost-effective method to continuously monitor low- and medium-importance assets. Using advanced vibration sensing technology, we can easily identify and repair unnecessary wear and tear on your assets, helping to extend the life of your assets. This also helps improve the overall efficiency of the equipment and reduces machine or process downtime.

Monitor assets – detect problems

For CbM and PdM, a number of different types of detection modes are available. Most applications involve current sensing, electromagnetic sensing, flow monitoring, and several other modes. Vibration detection is the most commonly used mode in CbM, and piezoelectric accelerometers are the most commonly used vibration sensors. In this section, we review how technological advances are driving the field of vibration sensors and how this impacts application decisions.

MEMS and Piezoelectric Accelerometers

Piezoelectric accelerometers are very high performance sensors, but achieving this performance requires many design trade-offs. For example, piezoelectric accelerometers are often used in wired installations because they consume too much power, can be bulky (especially triaxial sensors), and are expensive. Combined with all of these factors, it is not practical to use piezoelectric sensors throughout the plant, so they are generally only used on critical assets.

Until recently, MEMS accelerometers did not have sufficient bandwidth, were too noisy, and had g-ranges that only allowed monitoring of less critical assets. Recent advances in MEMS technology have overcome these limitations, enabling MEMS vibration sensors to monitor low-end assets as well as very important assets. Table 1 shows the important properties required for piezoelectric sensors and MEMS sensors in CbM applications. MEMS accelerometers are fast becoming the sensor of choice for many CbM applications due to their small size, years of battery operation, low cost, and performance comparable to piezoelectric sensors.

Compatible with MEMS and IEPE piezoelectric accelerometers, the CN0549 CbM development platform enables benchmarking between different sensor types.

Table 1. MEMS and Piezoelectric Accelerometers

Use of MEMS accelerometers in existing IEPE infrastructure

As shown in Table 1, MEMS accelerometers now offer competitive specifications and performance compared to piezoelectric sensors, but can they replace existing piezoelectric sensors? To make it easier for designers to evaluate and replace piezoelectric accelerometers with MEMS accelerometers, ADI has designed an interface that is compatible with the IEPE standard piezoelectric sensor interface actually used in CbM applications.

IEPE Sensor Interface and Mechanical Mounting (CN0532)

The CN0532 (shown in Figure 2) is an IEPE conversion circuit that allows MEMS accelerometers, like existing IEPE sensors, to seamlessly interface directly with the IEPE infrastructure.

Figure 2. CN0532 MEMS IEPE conversion circuit

Single-axis MEMS sensors typically have three output lines: power, ground, and acceleration output. The IEPE infrastructure only needs two: one line to ground and the other to carry power/signal. The current is transmitted to the sensor, and when the sensor detects vibration, the voltage is output by the same line.

Figure 3. Simplified schematic illustrating how MEMS sensors interface with existing IEPE infrastructure (power and data)

The CN0532 PCB is designed to be 90 mils thick to maintain the frequency response performance of the MEMS accelerometer given in the data sheet. The test unit is screw mounted and can be tested right out of the box. Mounting blocks, PCBs, solder pastes, etc. are all extensively characterized to ensure full bandwidth mechanical transfer capability, maximize visibility of various types of faults within the sensor bandwidth, and extend asset life by catching these faults. These solutions allow CbM designers to easily integrate MEMS accelerometers into their assets and seamlessly connect with existing piezoelectric infrastructure.

For high frequency vibration testing, the integrity of the mechanical signal path is very important. In other words, the vibration signal must not be attenuated (due to damping) or amplified (due to resonance) from the signal source to the sensor. As shown in Figure 4, an aluminum mounting block (EVAL-XLMOUNT1), four screw mounts, and a thick PCB ensure a flat mechanical response over the target frequency range. The IEPE reference design allows designers to easily replace piezoelectric sensors with MEMS sensors.

Figure 4. Vibration Measurement Test Setup: Attaching the EVAL-CN0532-EBZ Board to the Shaker Using the EVAL-XLMOUNT1 Aluminum Mounting Block

Figure 5. Frequency response of the EVAL-CN0532 compared to the frequency response given in the ADXL1002 data sheet

Vibrations to Bits – Data Transformation Integrity

Now, we know that MEMS sensors can be used in place of IEPE piezoelectric sensors. Also know how to easily fit them onto an asset while maintaining the performance given by the datasheet. It is important for the CbM development platform to be able to collect high-quality conversion data (whether based on MEMS or piezoelectric sensors) and then feed the data into the correct environment. Next, we look at how to acquire IEPE sensor data and maintain the highest data fidelity to develop the best CbM algorithm or machine learning algorithm. Our other CbM reference design, the CN0540, can help achieve this goal.

High-Fidelity 24-Bit Data Acquisition System for IEPE Sensors (CN0540)

Figure 6 shows a lab tested and validated IEPE DAQ signal chain. This reference design provides an optimized analog signal chain compatible with MEMS and piezoelectric accelerometers. ADI is not only focusing on MEMS accelerometer-based solutions. Note that piezoelectric accelerometers offer excellent performance and are widely used vibration sensors; therefore, piezoelectric accelerometers are important sensors for precision signal chain products.

The circuit shown in Figure 6 is a sensor-to-bit (data acquisition) signal chain suitable for IEPE sensors, consisting of current sources, input protection, level shifting and attenuation stages, third-order anti-aliasing filters, analog-to-digital converters (ADCs) ) driver and fully differential sigma-delta ADC. CbM system designers using piezoelectric accelerometers require a high-performance analog signal chain to achieve vibration data fidelity. Designers can evaluate the performance of the signal chain simply by connecting an IEPE sensor or CN0532 IEPE sensor directly to the CN0540 DAQ reference design. ADI has extensively tested this design and provides open source design files (schematics, layout files, bill of materials, etc.) for easy integration into end solutions.

The CN0540 IEPE data acquisition board is a tested and proven analog signal chain dedicated to acquiring IEPE sensor vibration data with better than 100 dB signal-to-noise ratio (SNR). Most solutions for interfacing with piezoelectric sensors on the market are AC coupled and do not have DC and subhertz measurement capabilities. The CN0540 is suitable for DC-coupled application scenarios where the DC component of the signal must be preserved, or the system must be guaranteed to respond to frequencies as low as 1 Hz or less.

The high-accuracy data acquisition reference design was tested with 2 MEMS sensors and 3 piezoelectric sensors, as shown in Table 2. As you can see from the table, the g-range, noise density, and bandwidth of each sensor vary widely, as does the price. Notably, piezoelectric sensors still have the best noise performance and vibration bandwidth.

For the CN0540, the system bandwidth is set to 54 kHz, and the signal chain noise performance is for a sensor capable of >100 dB dynamic range over this bandwidth, for example, a Piezotronics PCB 621B40 accelerometer can achieve 105 dB at 30 kHz. The CN0540 is designed to provide bandwidth and precision beyond the performance of current vibration sensors, ensuring that it does not become a hindrance when collecting high-performance vibration data. It is very easy to compare and benchmark MEMS and piezoelectric sensors on the same system. Whether working with MEMS sensors, piezoelectric sensors, or both, the CN0540 provides the best signal chain solution for data acquisition and processing, so it must be designed as an embedded solution.

When we say that a MEMS sensor offers comparable performance at a lower cost, the ADXL1002 has an SNR of 83 dB, but it costs more than 10 times less than a piezoelectric sensor. MEMS sensors can now replace all but the highest performance piezoelectric sensors at a low cost.

Figure 6. CN0540: Suitable for IEPE sensors for high-performance, wide-bandwidth, precision data acquisition

Table 2. MEMS and Piezoelectric Sensors and Their Corresponding Noise Density Measurements

Embedded Gateway

After the DAQ signal chain obtains high-fidelity vibration data, it must be processed and viewed in real-time and/or sent to a machine learning or cloud environment, which is the job of the embedded gateway.

Process vibration data locally in real time

Both embedded platforms are supported by Intel® (DE10-Nano) and Xilinx® (Cora Z7-07S), which includes support for all relevant HDLs, device drivers, software packages and applications. Each platform runs embedded ADI Kuiper Linux®, allowing you to display time and frequency domain data in real time, access real-time captured data via Ethernet, connect to popular data analysis tools such as MATLAB or Python, and even connect various Cloud computing instances (such as AWS and Azure). The embedded gateway can transmit 6.15 Mbps (256 kSPS × 24 bits) over Ethernet to your algorithm development tool of choice. Some key features of embedded gateways include:

● uIntel Terasic DE10-Nano

○ Dual-core Arm® Cortex®-A9 MP Core processor with 800 MHz neon™ frame media processing engine with double-precision floating-point unit (FPU)

○ 1 Gigabit Ethernet PHY with RJ45 connector

● uDigilent Cora Z7-07S (Xilinx)

○ 667 MHz Cortex-A9 processor tightly integrated with Xilinx FPGA

○ 512 MB DDR3 memory

○ USB and Ethernet connections

The IIO Oscilloscope (shown in Figure 7) is a free, open-source application installed with ADI Kuiper Linux that helps you quickly display time and frequency domain data. It is built on the Linux IIO framework and connects directly to ADI’s Linux device drivers, enabling device configuration, device data reading and visual display in one tool.

Figure 7. IIO scope showing FFT of 5 kHz pure tone

The ADI Kuiper Linux image also supports industry standard tools such as MATLAB and Python. IIO bindings were developed to stream data directly to these typical data analysis tools by using a connectivity layer that works with the IIO framework. Designers can use these powerful tools, combined with the IIO integration framework, to display and analyze data, develop algorithms, and perform hardware loop testing and other data processing techniques. Provides a complete example of how to transfer high-quality vibration data to a MATLAB or Python tool.

Predictive Maintenance Development Using CN0549

The development of machine learning (ML) algorithms for PdM applications generally involves five major steps, as shown in Figure 8. When performing predictive maintenance, regression models are often used, rather than classification models, to predict impending failures. The more training data you feed a predictive model, the better it performs. If only 10 minutes of vibration data is entered, it may not be possible to detect all operational characteristics, but if 10 hours of data is entered, the probability of detection is greatly increased, and if 10 days of data are collected, the performance of the model will be stronger.

Figure 8. Steps to develop a PdM application

Figure 9. CN0549 example use case

The CN0549 provides the data collection steps in an easy-to-use system where we can stream high-performance vibration data to the machine learning environment of choice.

MEMS IEPE sensors come with mechanical mounting blocks that allow seamless mounting of the MEMS sensor to an asset or shaker. Note that IEPE piezoelectric sensors can also be used with this system for easy installation into assets, shakers, etc. Before streaming the data to the data analysis tool, the sensor installation should be verified to ensure that there are no interfering resonances. This inspection can be done easily and in real time using an IIO oscilloscope. Once the system is ready, a use case can be defined, as shown in Figure 9, for example, a motor operating normally at 70% load. The high-quality vibration data can then be streamed to MATLAB or Python-based data analysis tools such as TensorFlow or PyTorch (and many others).

Through analysis, the characteristics and characteristics that define the health of the asset can be identified. Once you build a model that defines normal operating conditions, you can detect or simulate failures. Step 4 can be repeated to identify key characteristics that define the failure, thereby generating a model. Comparing fault data with data from a normally operating motor results in a predictive model.

The above is a brief overview of the machine learning process supported by the CbM development platform. Notably, the platform ensures the highest quality vibration data is delivered to the machine learning environment.

Part 2 of this article will detail the software stack, data flow, and development strategy, with examples using Python and MATLAB from the perspective of a data expert or machine learning algorithm developer. Additionally, software integrations will be outlined, along with on-premises and cloud-based development options.

The Links:   GP2500-SC41-24V QM500HA-H