top of page

Manufacturing Forensics and Self Optimization via AI

In order to make a system a smart system, streaming the data plays a very crucial role. We were able to stream data from three sensors, namely force sensor, accelerometer (vibration sensor), and acoustic emission sensor to any device that can access internet using web API and PI system. Security of machine data is crucial in this world and using a PI system helps to maintain the secure and reliable connection to the system enabling us for continuous insight of the acquired data and using it for applications not heard in the field of manufacturing before.
​

Objectives

The main objectives are:
  • Create a secure database for the sensor data and create a time synchronized historian to call the data.
  • Stream the real time data through a web API.
  • Apply analytics to the raw data and get useful insights into the data.

Highlights

Data Access

NI LabView is used for acquiring and storing the live sensor data of three sensors at the rate of 160  kHz collectively.

​

Data Storage

The cyber security of the data is important and to ascertain that data is stored securely, OSISoft's PI system is used as a one stop data storage solution. The data from labView is stored in a database using PI system.

​

Data Analysis

Dashboard created with an independent LabView program which displays not only the data stream but also data analytics such as FFT and spectrogram.

​

​

Capabilities and Demonstrations

Breakthrough: True potential of PI System has been explored. PI system is not designed for data analytics task and it has great capabilities as a plant integration system which means handling data at much slower rate. It is worth noting that ground breaking results were achieved in terms of writing and reading the amount of data to the PI System. We were able to write 160 kHz on the PI System from the sensors. Moreover, the entire setup is capable of reading up to 160 kHz of data. This is a ground breaking achievement since such rates being achieved using a PI system is not known.

​

Scalability: The setup that is developed is scalable. Endless number of channels can be added to the setup and data from those channels can be monitored, stored, and analyzed. However with additional channels, computational needs also increases. For three channels, the windows server with Intel I7 (8th Gen) with 24 GB of RAM was used. 

​

Data Analysis: The program and the setup is capable to perform data analysis. As of now, the setup is able to perform basic data analytics by plotting FFT and spectrograms. The system is very much capable to perform advanced data analysis such as wavelet transformations which would be covered in future.

​

​

In order to initiate the system, PI Connector Relay is activated. Once that is done, the connection between PI Connector for OMF and OMF Application Data is made, with PI Connector Relay bridging the both. This is done using PI Data Collection Manager.

Screenshot (14).png
PI Data Collection Manager

Essentially, now the connection between NI DAQ system (inclusive of NI LabView) and OSISoft PI System has been established. After creating a LabView program, one can explore the capabilities of the PI System as well as LabView. NI LabView and OSISoft PI are well integrated with each other. Multiple virtual instruments (VIs) can be obtained from LabView Connectivity Toolkit which is used to write a LabView program and integrate it effectively with OSISoft PI System.

Screenshot (13).png
LabView UI for Writing Data on PI System

From the UI, one can start writing data onto the PI System by clicking the run button. The initiation and ending of the program needs to be done manually using the UI shown above.

Screenshot (15).png
LabView UI for Reading Data from PI System

The dashboard created with the help of a LabView program displays signals from all the channels connected with the DAQ. The dashboard also displays FFT plots and Spectrogram plots for each channel. These plots are generated from computation performed in the background in a LabView program. One can adjust the sampling rate manually. Just like initiating writing, one needs to manually initiate and end the reading program using the UI integrated with the dashboard. It is worth noting that due to computational limitations, the signal plotting and analytics plotting lags and a true real time data streaming has not been realized. However, this can be overcome by overhauling the system.

bottom of page