Multinode High-Precision Synchronized Inertial Measurement System

"We developed this system in just four months by using LabVIEW and CompactRIO, and succeeded in performing our measurements in the world’s largest vibration experiment using an 18-story reduced scale test structure for quantifying the collapse margin of steel construction high-rise buildings."

- Kazuyoshi Takeda, WP Planning & Design Department, Wearable Products Operations Division, Seiko Epson Corporation

The Challenge:

Developing a large-scale, reliable, and high-performance system to measure the synchronization of more than 200 inertial measurement units (IMUs) in milliseconds.

The Solution:

Using LabVIEW software and CompactRIO hardware to build a multinode high-precision synchronized inertial measurement system in four months.

Corporate Overview

Seiko Epson's sensing system business has produced highly precise, stable, reliable sensors that are also ultra-compact, by using its QMEMS quartz microfabrication technology. This technology is also behind the company’s core product: wristwatches. Seiko Epson creates various proposals for maintaining social infrastructure and contributes to society in many ways, from monitoring bridge deterioration and building vibrations to preventive maintenance for faults in manufacturing devices.


QMEMS combines quartz, a crystal material with excellent characteristics such as high stability and high precision, and MEMS, a microfabrication technology. Seiko Epson applies precision microprocessing based on quartz material to a MEMS made of semiconductor materials to create quartz devices known as QMEMS, which offer high performance in a compact package. QMEMS is a registered trademark of Seiko Epson Corporation.



We worked on Subproject #2 of the Special Project for Reducing Vulnerability for Urban Mega Earthquake Disasters called Maintenance and Recovery of Functionality in Urban Infrastructures (a five-year plan from 2012-2016), commissioned by the Ministry of Education, Culture, Sports, Science, and Technology, Japan (MEXT). For this project, we used an 18-story reduced scale test structure to conduct the world’s largest scale shake table experiments for quantifying the collapse margin of steel construction high-rise buildings. As a new approach to structural health monitoring, the Shimizu Institute of Technology incorporated Epson IMUs in sensors to estimate damage to areas such as the structural framing at a material level. In a short period of time, we developed a multinode system using more than 200 IMUs to perform high-precision synchronization and measurement. We expect to use this system in three experiments under this subproject, including the experiment referred to in this paper.


There were two important lessons learned from the 2011 Tohoku earthquake. We need:
• Measures to address seismic motion that exceeds anticipated levels
• A continued and prompt recovery of business and lifestyle


Through this subproject, we aim to achieve the following with regard to these two lessons learned:
・ Quantify collapse margin of high-rise buildings
・ Monitor and promptly assess the condition of buildings following an earthquake


System Requirements

The main requirements for the multinode high-precision synchronized inertial measurement system are:
• Perform high-precision synchronization of more than 200 Epson IMUs in the order of milliseconds.
• Perform measurements for up to 10 minutes at a sampling rate of 500 samples per second (sps).
• Establish robust sensors, a means of communication, and a measurement system since the system would be installed within the test structure for the collapse tests.
• Succeed even when 3G acceleration is applied.
• Meet the delivery date (development period was four-months long).
• Ensure extremely high reliability. (Several hundred million yen and several years would be invested jointly by industry, government, and academia on preparations for the experiment, including preparing the test structure and test environment. Furthermore, the experiment would be a destructive test, so retesting would not be possible. Accordingly, there would be no margin for error in measurement.)


Turning to NI for Solutions

Developing a measurement system from scratch that could meet the requirements outlined above would require years of time, and we would not meet the requested delivery date. We turned to NI solutions, which include completed hardware products that are highly reliable, have a proven track record for measurement and control applications, and also provide architecture fully suited to large-scale system development.


We created a distributed processing system using CompactRIO, and we built subcontrollers by connecting up to 48 IMUs to each single CompactRIO unit. We used these subcontrollers to build a system that could simultaneously collect data at 500 sps from 48 nodes x 6-axis sensors by processing at a high speed using CompactRIO FPGA circuits. This system comprised five subcontrollers and could connect up to 240 IMUs. We could also increase the number of connected IMUs by increasing the number of subcontrollers, which made this a highly scalable system.


Due to the IMU wiring constraints, we needed to install the CompactRIO device on the vibration test structure. The robust CompactRIO guaranteed operation even when we applied 5G acceleration.


We connected CompactRIO and the IMUs using controller area network (CAN) interfaces. CAN interfaces connected the IMUs to the CompactRIO units because they needed only minimal wiring to connect the IMUs, enabled high-precision synchronization, and are highly resistant to external impact and noise. We used pulse signals to synchronize the CompactRIO units with minimal elements of delay and jitter. We also used an optical cable to connect CompactRIO units positioned further away from each other with low loss.


The master controller and subcontrollers used a one-to-many connection. We used a suitable Ethernet cable to connect the master controller (installed in the measurement room) to the subcontrollers (installed within the test structure) because the distance between these was several tens of meters. The biggest concern we had for the collapse of the test structure was that the Ethernet cable connecting the structure to the outer world would become ruptured. Once measurements began under this system, each subcontroller could perform measurements independently, without instruction from the master controller, and the measurement data was stored in the non-volatile memory within the subcontroller. Accordingly, we ensured that measurements could continue to be taken, and measurement data could be stored, even if the Ethernet cable became ruptured.


System Overview

Figure 2 shows the hardware configuration, including:
• The master controller (PC) connects via Ethernet to five subcontrollers (cRIO-9082).
• Four 2-port, high-speed CAN modules (NI-9853) connect to the subcontrollers, and up to six IMUs connect to each CAN port through a circuit that combines the power supplied to the IMUs.
• The subcontrollers consist of one synchronization master with the rest as synchronization slaves. The synchronization master outputs synchronizing signals from a 4-ch digital I/O module (NI-9402), and the synchronization slaves input these signals using the same module. The synchronizing signals are converted to optical signals along the way by an electrical-to-optical signal converter.


Figure 3 shows the software processing sequence during measurement, including:
1) LabVIEW running on the master controller sends commands to each subcontroller real-time OS (RTOS) to begin measurement.
2) The synchronization master RTOS starts the FPGA, and the FPGA sends synchronized signals at sampling intervals (1/500 sps = 2 milliseconds).
3) Each subcontroller FPGA is synchronized by the synchronizing signals and outputs a measurement synchronization message to each CAN port.
4) Each IMU is synchronized by the measurement synchronization message and outputs the measurement data.
5) Each subcontroller FPGA receives the measurement data within the sampling interval and sends this in bulk to the RTOS.
6) The RTOS saves the data from the FPGA to a file for each IMU.
7) Each subcontroller automatically finishes measuring when the measurement time has elapsed and notifies the master controller that measurement has ended.
8) After the master controller has received notifications from all subcontrollers that measurement has ended, it collects the measurement data from each of the subcontrollers.




We could not have achieved the superior performance of the measurement system without an NI measurement system or Epson IMUs. In recent years, the industrial and commercial fields have been demanding improved precision and stability of various kinds of sensors, while also requiring them to be more compact and consume less energy to make them easier to embed in products. Seiko Epson, as the quartz device industry leader, has adopted a unique approach to developing IMUs. Quartz gyro (angular velocity) sensors fortified with the company’s signature QMEMS quartz microfabrication technology are fused with semiconductor technology and knowledge gleaned from location information devices such as GPS. The resulting IMUs brought to market by Seiko Epson achieve the world’s smallest class of external dimensions (24x24x10 mm) and lowest power consumption (30 mA when the operating power supply voltage is 3.3 V) while also boasting highly precise, highly stable measurement performance.


These IMUs are equipped with 6-axis sensors that consist of high-precision, high-stability 3-axis gyrosensors and 3-axis accelerometers. In terms of measurement performance of angular velocity, they achieve angle random walk performance of 0.2 deg/h and gyro bias stability of 6 deg/h. We also brought the IMU sensor units to market. We packaged these in cases with waterproof and dustproof properties equivalent to IP67 and equipped with a CANopen-compliant communication protocol.


These IMUs are suitable for applications in industrial and commercial fields, including for analysis and control of inertial movement, motion analysis and control, moving body control, vibration control and stabilization, and navigation systems. They are more compact, lightweight, and energy efficient than conventional IMUs, which makes them more flexible when designing the products into which they will be embedded. Furthermore, users can now embed them in an even wider variety of products in industrial and commercial fields, which had been difficult to achieve previously.


This experiment demanded high precision and highly stable sensor performance of a level that was hard to reach with consumer-oriented sensors. Furthermore, we installed more than 200 IMUs within the test structure, so they needed to be compact, energy efficient, easy to handle, easy to wire, and deliver reliable communications.


Benefits of Implementation

We achieved the following benefits by adopting NI solutions:
1) By using NI solutions, which include NI’s highly reliable established hardware, strong training programs, and software development support services, a LabVIEW novice developed PC and CompactRIO software to complete this system in just four man-months. A team of three people total developed the system itself, with one working on hardware development, one on software development, and one in charge of assessments. In particular, the strong training system and software development support is unrivalled by other companies. Without this supportive environment, we could not have completed the system in just four months. If we had tried to develop this system from the hardware up without utilizing NI solutions, we estimate that it would have taken two to three years and hundreds of millions of yen in development expenses to complete the system.


2) The excellent processing capabilities of CompactRIO enabled the CAN synchronization messages to each IMU to be achieved with synchronization precision of approximately 3.5 milliseconds. Furthermore, we could sample measurement data from the 6-axis sensors of up to 48 IMUs per CompactRIO unit (total 288 axes) at 500 sps, achieving a system capable of receiving, converting, and saving to file data of approximately 6 Mbit/s. We largely attribute this superior performance to the CompactRIO FPGA. Furthermore, we could develop the FPGA circuits in such a short time thanks to the compatibility of NI solutions for developing FPGA circuits in a common development environment with LabVIEW.


3) The excellent reliability of CompactRIO empowered us to conduct vibration experiments over three days on 152 IMUs at 500 sps, to complete measurements without loss of measurement data and without any system trouble. The test structure collapsed at the completion of the vibration experiments, rupturing the Ethernet cables. However, CompactRIO independently carried out measurements to the end, and we safely retrieved the measurement data after collecting the CompactRIO units.


Furthermore, by building this system, we could participate in Subproject #2 of the Special Project for Reducing Vulnerability for Urban Mega Earthquake Disasters commissioned by MEXT and contribute to the Shimizu Institute of Technology’s research on structural health monitoring. We developed this system in just four months by using LabVIEW and CompactRIO, and succeeded in performing our measurements in the world’s largest vibration experiment using an 18-story reduced scale test structure for quantifying the collapse margin of steel construction high-rise buildings. We hope that this research helps construct mechanisms for the prompt assessment of the structural integrity of urban infrastructure immediately following an earthquake, leading to the swift recovery of business and lifestyle.


The test structure was 18 stories high. We installed eight Epson IMUs on each floor (152 total) and a subcontroller containing CompactRIO on every few floors, with five total throughout the overall structure.


Even under these conditions, our system continued to successfully perform measurements.


Figure 8 shows an example of data measured using this system and an example of the estimated damage results calculated by the Shimizu Institute of Technology. The vertical direction indicates the building’s 18 stories. The distribution of damage to the Y1 and Y2 structural surfaces of the test structure is indicated by white for no damage and gray for damaged. The ruptured location of the beam-end flange corresponded roughly with the estimated results, indicating that it was possible to estimate the distribution of damaged locations within the test structure.


Summary and Future Outlook

Moving forward, we would like to further expand the features listed below based on this system, and create one that is even easier to use for measuring building structures.
• Display measurement data in real time during measurement, and shorten processing and collection times for post-measurement data
• Improve flexibility for installing IMUs (expand length of communication-enabled sensor cables)
• Connect other Epson QMEMS sensors (vibrometers, angle meters)
• System control functionality using wireless communication



[2] (in Japanese)


Author Information:

Kazuyoshi Takeda
WP Planning & Design Department, Wearable Products Operations Division, Seiko Epson Corporation

Figure 1. Subproject Objective
Figure 2. Hardware Configuration
Figure 3. Software Processing Sequence During Measurement
Figure 5. Epson M-G550PC IMU Sensor Unit
Figure 6. (1) Test Structure Installed on the Shake Table in the Experiment, (2) Epson IMUs Installed Within the Structure, (3) A Subcontroller Installed Within the Structure
Figure 7. Test Structure After It Collapsed at the Completion of Vibration Testing
Figure 8. (1) Example of Measurement Data, (2) Example of Estimated Material Damage Results
Figure 4. LabVIEW Screenshots Taken During Measurement (left),
When Checking IMU Connections (right)