Understanding InsightCM™ Continuous Monitoring Device Performance

Overview

NI InsightCM is a flexible application that allows the user to change a range of settings to capture data that can help indicate a failure mode. However, for assets that require sampling at very high frequencies, you may inadvertently configure your device in a way that leads to unstable performance and reboots. If the device reboots 10 times within one hour, it will report an Invalid Configuration. Settings that affect device performance include file length, sample rate, channel count, and channel integration. NI recommends keeping CPU and memory usage below 80% to ensure stable performance, which you can monitor on the Device Health Dashboard.

Contents

Choosing the Right Configuration

When configuring NI monitoring device hardware, you should consider the characteristics of the asset you are monitoring before choosing the ideal configuration.  

First, you should select a sample rate that is at least 2.56 times greater than the maximum vibration frequency you would like to measure, also known as the FMax. Also consider how many Lines of Resolution you require, which is related to FMax by the following equation:

Where dF is the frequency resolution, or the FFT “bin size.” dF is the inverse of File Size in seconds:

In certain scenarios, you may need to properly balance the need for longer file lengths and higher sample rates on the same device. Slow speed equipment may require a long file length to capture enough revolutions, while high speed equipment often requires high sample rates to capture the frequency content of interest. If your equipment requires a high sample rate for some sensors and a long file length for others, you may need to use two separate devices to capture the data.

 

Back to top

Benchmark Setup

To demonstrate the effect of the sample rate setting on device performance, multiple configurations were benchmarked. The following three configurations were tested on a CMS-9068 and CMS-9036 connected by Gigabit Ethernet to a server with InsightCM’s recommended specifications:

  • Configuration 1: 1 Tachometer, 3 Accelerometers
  • Configuration 2: 1 Tachometer, 9 Accelerometers
  • Configuration 3: 1 Tachometer, 21 Accelerometers

Sample Rates of 12.8, 25.6, 51.2, and 102.4 kS/s were tested with the default accelerometer and tachometer feature set. A file length of 4 seconds and pre-trigger length of 1 second was used in all test configurations. Data Sets were collected once an hour, and Trend Data was collected every 15 minutes. No Delta EU triggers were used for data collection.

The CMS-9065 was not tested because it is the 4-slot variant of the CMS-9068, but it should perform similarly due to its identical internal components.

The data below should only be used as a reference for baseline performance, due to the many factors that can affect performance. Increased network latency, more connected InsightCM devices, increasing data collection frequency, and poor server performance will all negatively affect device performance. For instance, adding a Delta EU Trigger will increase your device's memory usage whenever a given feature is changing rapidly.

 

Back to top

Results

Generally, memory usage will be much higher than CPU usage on InsightCM devices, so only memory usage was displayed in the charts below.

Figure 1: Device Configuration 1. With only 3 accelerometers, both devices ran well below the recommended memory threshold of 80%.

 

Figure 2: Device Configuration 2. The CMS-9068 was unable to collect data with this configuration at 102.4 kS/s.

 

Figure 3: Device Configuration 3. Neither the CMS-9068 nor the CMS-9036 could collect data at 102.4 kS/s in this configuration. The CMS-9068 was also unable to sustain a sample rate of 51.2 kS/s.

 

To illustrate the difference changing the file length settings can make, the above test was repeated with 1 second file lengths, instead of 4 seconds:

Figure 4: Configuration 3 with 1 second file lengths compared to 4 second file lengths.  

 

Lastly, the CMS-9036 was tested with single and double integration enabled on all 21 accelerometer channels with 4 second file lengths:

Figure 5: Various CMS-9036 channel configurations with a sample rate of 25.6 kS/s and file lengths of 4 seconds.

 

Adding single and double integration to a device’s channels has a significant effect on performance, due to the inherent computational overhead.

 

Back to top

Conclusion

As stated before, there are several factors beyond sample rate and channel count that will affect device performance, so the data in this White Paper should be used as a baseline. If you are dealing with device performance issues, consider making the following changes:

  • Decrease the device sample rate
  • Shorten the file length setting for the data groups assigned to the device
  • Remove integrated features, which consume additional memory and computational resources on the device.
  • If the device is collecting data sets multiple times each minute, change the trigger settings to collect data less often.

Choosing the right settings to change is dependent on the assets you are monitoring and the data you care most about.

 

Back to top