“Government initiatives, such as Industry 4.0 in Germany and Made in China 2025 in China, are accelerating the trend towards pervasive network automation in manufacturing. In addition, smart sensor systems are increasing automation, providing more data to monitor and control the production process.
“
By Richard Anslow and Dara O’Sullivan of Analog Devices
Condition-based monitoring is essential for enabling Industry 4.0
Government initiatives, such as Industry 4.0 in Germany and Made in China 2025 in China, are accelerating the trend towards pervasive network automation in manufacturing. In addition, smart sensor systems are increasing automation, providing more data to monitor and control the production process. In particular, Made in China 2025 aims to rapidly develop high-tech industries including electric vehicles, next-generation information technology (IT) and telecommunications, advanced robotics and artificial intelligence. With more advanced systems, more advanced methods are needed to ensure system reliability.
Condition-based monitoring of robots and rotating machines, such as turbines, fans, pumps, and motors, records real-time data about machine health and performance for targeted predictive maintenance and optimal control. Targeted predictive maintenance early in a machine’s life cycle can reduce the risk of production downtime, resulting in increased reliability, significant cost savings and increased plant productivity.
How to implement a condition-based wired monitoring solution?
To implement condition-based monitoring of industrial machines, a range of sensor data can be leveraged, such as electrical measurements, vibration, temperature, oil quality, acoustics, and process measurements such as flow and pressure. However, vibration measurement is by far the most common as it is the most reliable indicator of mechanical problems, such as imbalances and bearing failures. This article focuses on the application of vibration sensing, but the method is equally applicable to data from other sensors.
This transmission means sending sensor data from sensing nodes to a host controller or cloud highly dependent application. In many applications, some local data is processed at the end node, and aggregated data is then sent wirelessly to a network gateway, or directly over a cellular link to the cloud or analytics server. In these cases, the amount of data transferred is usually quite low, and because the end node is battery powered, it is usually required to keep power consumption low. In other applications, raw sensor data transfer is required. For example, data from multiple sensors may need to be adjusted and fused prior to analysis. In applications that use data for real-time control, raw data transfer is also required. In these applications, a wired interface is more likely to be used as the data transfer solution.
Figure 1. Options for implementing a reliable, highly integrated, wired MEMS accelerometer-based solution for condition monitoring.
CbM for industrial applications can use ADI’s optimized microelectromechanical systems (MEMS) accelerometers, low-power microcontrollers, and wired iCoupler® isolated interface signal chains to extract, adjust, and adjust machine health data from remote CbM slaves. It is reliably transmitted back to the main controller for analysis. Over time, machine health data can be used to create software-based models to identify changes in machine behavior and proactively maintain machine health. In some applications, such as CNC machine tools, the data can also be used to optimize system performance in real time.
Challenges in implementing a wired CbM interface include EMC robustness when operating over long cables, data integrity when transmitting at high baud rates (for real-time streaming of CbM data), and communication physical layer/protocol mismatches. Analog Devices’ signal chain and system-level expertise provides several possible options for implementing a wired CbM interface.
This article is divided into two parts. The first part introduces ADI’s wired interface solution, which helps customers shorten design cycle and test time, and bring industrial CbM solutions to market faster. The next article focuses on detailed physical layer design considerations, including the master controller and the wired CbM slave controller.
Wired CbM Design and Implementation
Designing and deploying a condition-based wired monitoring solution requires consideration of multiple system performance factors and trade-offs.
First, in choosing the right MEMS accelerometer, the type of fault that needs to be measured must be considered to select the appropriate bandwidth and noise performance MEMS to meet the requirements of the system. Edge node processing requires careful matching of selected processors to ensure maximum system flexibility.
Second, the design of a wired CbM system requires careful selection of appropriate wired communication protocols and physical layers to achieve high-speed real-time data streaming. Implementing a wired interface requires careful consideration of EMC performance, data transmission cables, connectors, and power delivery over cables.
Choosing the Right MEMS Accelerometer
Choosing the right MEMS vibration sensor covers several aspects:
number of axes
The number of axes monitored is usually a function of the type of fault and the placement of the sensors. If it is obvious that the fault involves a capstan shaft and there is a clear transmission path on this shaft, then a single-axis sensor is sufficient. Triaxial sensing is useful for faults containing energy in multiple axes or faults where the path of the fault energy is not well defined.
Fault type
The type of fault being monitored has a significant impact on sensor selection. The noise density and bandwidth of the sensor are important metrics in this regard, as they determine the vibration level and frequency range that can be reliably extracted. For example, unbalance and misalignment faults in low-speed machines may require a low noise density sensor but with fairly low bandwidth requirements, while gear fault detection requires a sensor with both low noise density and high bandwidth.
performance requirements
In addition to failure types, it is important to understand the performance requirements of CbM. Implementing alarm detection for status indicators for basic traffic light types requires complex predictions with varying levels of performance. This obviously applies to the analytics and algorithms being deployed, but also affects sensor selection. The higher the sensor’s performance level in terms of bandwidth, noise density, and linearity, the greater the analytical capabilities.
Choosing the Right Signal Processing
Design considerations include:
Accelerometer output
The output of an accelerometer is typically an analog or serial digital signal, usually SPI. Analog output sensors will require a digital conversion stage and also some signal conditioning. This can be a discrete ADC supporting preamp conditioning, or an embedded ADC in a microcontroller.
Edge Node Processing Requirements
To offload the data link and/or the central controller/server, some basic FFT or signal processing algorithms may be required on the edge nodes.
Data Transfer Protocol Requirements
The output of the ADC or sensor is usually an SPI interface. It does not by itself provide any mechanism for implementing data integrity checks, determining timestamps, mixing data from different sensors, etc. It is useful to encapsulate sensor data in high-level protocols at edge nodes prior to transmission. This can improve the robustness and flexibility of the sensor interface, but requires proper processing and encapsulation of data streams at edge nodes.
For more information, refer to the Analog Dialogue article “Choosing the Best MEMS Accelerometer for Your Application”.
Porting accelerometer output to wired communication bus
As mentioned earlier, the output of an accelerometer is typically an analog or serial digital signal, usually SPI. The SPI output can be handled locally (allowing for protocol flexibility) and then added to the physical layer interface, or ported directly to the physical layer.
SPI is an unbalanced single-ended serial interface used for short distance communication. To port SPI directly to the physical layer over longer distances requires the use of RS-485 line drivers and receivers. RS-485 signaling is balanced differential, inherently immune to interference, and robust over long cable lengths.
There are some challenges when using SPI over longer distances between SPI master and slave. SPI is inherently synchronous, with a clock (SCLK) initiated by the SPI master. The SPI data lines—Master Out Slave In (MOSI) and Master In Slave Out (MISO)—are synchronized to SCLK, which is achievable over short distances. SPI also has an active, low-enable chip select (CS) signal that allows individual slave addressing if required.
To restore synchronization between the master and slave, the slave’s clock signal can be fed back to the master, or the clock phase shift can be used to compensate for the master’s cable delay. The phase shift of the clock must match the total delay of the system. AN-1397 provides implementation details of delay compensation for the host microcontroller.
wired communication physical layer
When communicating over long distances, a robust and reliable physical layer is required. As mentioned earlier, RS-485 signal transmission is a balanced differential transmission, which is inherently anti-interference. System noise is equally coupled to each conductor in an RS-485 twisted pair cable. The emission of one signal is the opposite of the other, and the electromagnetic fields coupled to the RS-485 bus cancel each other out. This reduces the electromagnetic interference (EMI) of the system. Some additional key advantages that make RS-485 ideal for CbM systems include:
• Higher data rates, up to 50 Mbps for short cable lengths (less than 100 meters)
• Cable lengths up to 1000 meters at lower data rates
• Full/half duplex RS-485 and RS-422 multiple driver/receiver pairs can convert bidirectional SPI to RS-485 bus signals with minimal components
• Wide common-mode input range allows for ground potential differences between master and slave
• EMC performance of wired interface
Communication networks can be affected by hazards such as large common-mode noise, ground potential differences, and high-voltage transients when traveling in long cables.
Conducted and radiated noise sources can affect communication reliability over 100 meters of cable length. Using Analog Devices’ iCoupler chip-scale transformer isolation technology can improve immunity to these noise sources. AN-1398 outlines the resistance to common industrial transients that can be achieved with iCoupler technology.
In a factory automation environment, system designers often have no control over the electrical devices that provide the communication network. The best practice is to assume that there is a ground potential difference. In motion control systems, ground potential differences of hundreds of volts can occur. RS-485 communication nodes require galvanically isolated power and data lines to operate reliably in these environments. Signal and isoPower isolation devices offer maximum continuous operating voltages up to 600 V (base) or 353 V (enhanced) peak. Basic insulation supports reliable communication in the presence of large ground potential differences. Reinforced insulation protects operators from electric shock in the plant area.
In wired communication networks, exposed connectors and cables can be subjected to many severe high voltage transients. System-level IEC 61800-3 standard related to EMC immunity requirements for variable-speed electric drive systems, requiring minimum ±4 kV (contact)/±8 kV (air) IEC 61000-4-2 ESD protection. ADI’s next-generation RS-485 transceiver provides greater than ±8 kV (contact)/±8 kV (air) IEC 61000-4-2 ESD protection.
Phantom power on data lines
Distributing power and data lines between the main controller and remote CbM sensor nodes requires innovative solutions to reduce cable costs. Converging data and power lines on a single twisted pair means significant savings in system cost, as well as a smaller printed circuit board (PCB) connector solution for end sensor node locations where space is limited.
Power and data are distributed over the twisted pair through an Inductor-capacitor network. High frequency data is coupled to the data line through series capacitors while protecting the RS-485 transceiver from the DC bus voltage. The power supply on the master controller is connected to the data line through an inductor, which is then filtered using the CbM at the far end of the cable from the inductor on the sensor node.
The inductances at both ends of the cable are well matched to avoid differential mode noise, and the self-resonant frequency should be at least 10 MHz to avoid interfering with the real-time burst mode of ADI’s next-generation vibration measurement system. Note that power and data coupling solutions must be added to data lines that do not require DC data content, such as MOSI or MISO to RS-485 extensions.
Recommended solutions and performance trade-offs
Based on the proposed design considerations, the following components provide the best path to a robust wired industrial vibration measurement solution.
• ADcmXL3021, Wide Bandwidth, Low Noise, Triaxial Vibration Sensor
• ADuM5401/ADuM5402, quad, 2.5 kV isolator with integrated DC/DC converter
• ADM3066E, 50 Mbps half-duplex RS-485 transceiver
• ADM4168E, 30 Mbps dual channel RS-422 transceiver
• LTC2858-1, 20 Mbps full-duplex RS-485 transceiver
• ADP7104, 20 V, 500 mA, low noise CMOS LDO regulator
Recommended solution
The ADcmXL3021 MEMS accelerometer is common to all three solutions. This accelerometer has ultra-low noise density (25 µg/√Hz) and supports excellent resolution. The ADcmXL3021 also has a wide bandwidth (from DC all the way to 10 kHz, 5% flatness) to track key vibration characteristics on many machine platforms. The ADcmXL3021 provides customers with a mechanically optimized aluminum package that provides stable coupling to integrated MEMS sensors over a wide frequency range. This ensures that vibration signatures obtained from the device under test can be reliably extracted and conditioned.
The ADcmXL3021 can provide an SPI output and can interface directly with RS-485/RS-422 devices or through microprocessor and/or iCoupler signal and power isolation, as shown in Figure 1. For real-time monitoring of vibration signatures on industrial equipment, the ADcmXL3021 offers a real-time streaming mode that operates at approximately 12 Mbps SPI.
In order to connect the live streaming SPI mode to the RS-485 bus, components with excellent data rates must be selected.
The ADM3066E/ADM4168E/LTC2858-1 RS-485/RS-422 transceivers all operate at data rates of 20 Mbps and above.
Figure 2. LTC4332 SPI expansion interface helps save cabling costs.
For Option 1 and Option 2 shown in Figure 1 (which can be connected directly to RS-485 via SPI), the ADM3066E and ADM4168E provide a robust interface for SPI 3 receive, 1 transmit (3+1) at the slave vibration sensor node configuration. The SPI CS receive signal is implemented using the ADM3066E, SPI CLK and MOSI, and the MISO signal is implemented using the ADM4168E. When operating in live streaming mode, the ADcmXL3021 sends an interrupt signal to the host microcontroller to flag when a new burst of data can be captured. The interrupt signal (/BUSY) can also be transmitted to the host using the ADM4168E.
The complete solution consists of three signals (MOSI, CS, CLK) sent from the host to the ADcmXL3021 and two signals (MISO, /BUSY) sent back to the host from the ADcmXL3021. A 5× single-ended signal can be converted to a differential signal using only two components, the ADM4168E and ADM3066E. Differential signals can be converted using RJ50 connectors and plugs, both of which take up nearly the same PCB area compared to industry standard RJ45 Ethernet connectors. The ADM3066E and ADM4168E transceivers provide greater than ±8 kV (contact)/±8 kV (air) IEC 61000-4-2 ESD protection, providing the necessary reliability when connected directly to a wired cable interface.
For option 3, the microcontroller can preprocess the ADcmXL3021 SPI output or perform protocol conversion between SPI and other serial interfaces such as UART. UART is an asynchronous protocol commonly used in RS-485 interfaces. The UART consists of transmit and receive signals and transmit enable signals, all of which can be connected directly to a full-duplex RS-485 transceiver such as the LTC2858-1. In full-duplex mode, the LTC2858-1 allows simultaneous bidirectional data transfer, which matches the requirements of SPI bidirectional data transfer. This microcontroller can handle the conversion of synchronous SPI to asynchronous UART protocol.
The ADuM5401/ADum5402 are the industry’s smallest signal and power isolation devices. They contain an integrated DC/DC converter that provides up to 500 mW of regulated isolated power at 5.0 V or 3.3 V (5.0V input supply).
In Figure 1, Option 2 includes the ADuM5401, which takes 5 V DC from the data bus and then provides 3 V isolated power to the ADcmXL3021. The ADuM5401 also includes four signal isolation channels in a configuration that supports 3+1 SPI isolation.
Option 3 in Figure 1 includes the ADuM5402, which is similar to the ADuM5401. The key difference is that the ADuM5402 provides 2 transmit and 2 receive digitally isolated channels.
As mentioned earlier, the ADuM5401/ADuM5402 can improve the EMC immunity of the wired CbM interface and protect the ADcmXL3021 from high voltage interference and ground potential differences on the RS-485 cable interface.
Performance trade-offs
Table 1 compares the three solutions using a number of key metrics, including design flexibility, PCB area, solution cost, complexity, and EMC performance.
Integrating a microcontroller in the CbM sensor node will increase design flexibility, but will increase PCB area and software complexity. Since the main CbM node will have a processor, this means that option 3 in Figure 1 will essentially be a dual microcontroller system that is up and running compared to a single microcontroller on the main CbM node The speed will be slower.
Option 1 and Option 2 offer less design flexibility, but offer a faster path to deployment because they support low-complexity and transparent SPI integration over the RS-485 link. Option 1 and Option 2 can also use a smaller PCB than Option 3, which requires additional PCB area for the microcontroller and associated circuitry (eg, a clock oscillator and several passive components).
Adding iCoupler signal and power isolation to Option 2 and Option 3 requires minimal PCB area and can improve EMC performance (beyond what can be achieved using on-chip protection of RS-485/RS-422 transceivers).
Table 1. Comparison of trade-offs between CbM options
|
Solution options |
Design flexibility |
PCB area |
Solution cost |
complexity/integration |
EMC performance |
|
1 |
Low |
Low |
Low |
Low |
middle |
|
2 |
Low |
low/medium |
low/medium |
Low |
high |
|
3 |
high |
middle |
middle |
middle |
high |
Solutions for lower data rates
For wired applications operating at lower data rates (less than 2 Mbps), the LTC4332SPI expander provides an alternative for hardening the SPI link between master and slave sensor nodes. The LTC4332 can transmit SPI data, including interrupt signals over two twisted pair wires. This solution offers significant cost savings as it saves up to 50% of bus cabling compared to standard solutions.
About the Author
Richard Anslow is a systems applications engineer in the Interconnected Motion and Robotics team in the Automation and Energy business unit at Analog Devices. His areas of expertise are condition-based monitoring and industrial communication design. He holds a Bachelor of Engineering and a Master of Engineering from the University of Limerick, Ireland. Contact information:[email protected]
