Patrick Mannion
With the increase of sensory nodes, data does not take a long time to form data streams, then data streams are formed, and data lakes are quickly formed. Engineers responsible for managing machinery and systems will soon be overwhelmed by data.
The universal data platform may not be of much use. Therefore, analysis software needs to "think like an engineer." In other words, it needs a framework to handle the engineer's priorities for maintenance, fault prediction, and troubleshooting.
This is what Flutura wanted to achieve when creating the Cerebra platform. Cerebra's key differentiator advantage is that it uses the engineer's thinking model to provide data.
"With Cerebra, we can turn data lakes into an engineer's framework with algorithms that can detect common 'accidents' that can lead to [system] failures," said Derick Jose, chief product officer at Flutura.
Engineer's thinking model
For anyone who has attended engineering training, Jose's description of the thought model may sound familiar, and Cerebra uses this model to quickly present data about the asset.
â—? What process does the asset follow? (Explain the operation intent and parameters.)
â— How can the assets fail? (Define the failure mode.)
â— What is the nature of the interventions in the past 90 days? (Track maintenance and repair history.)
â—? How about the operating environment? (Are you in Siberia or Texas?)
â— How does the alarm frequency trigger? How often does the alarm unattended? (Analysis of historical behavior.)
Jose said: "Unless you reflect this worldview on data products, it is very difficult for information users to handle it and get effective insights from it."
Although the model can be applied to any IoT application, Flutura currently focuses on the energy and engineering industries (Figure 1).
Figure 1. Flutura's Cerebra platform uses a deep learning algorithm to map data to the engineer's perspective. (Source: Flutura)
Map data to thinking models
The process of turning a lake of raw data into an engineer's thinking model is applying for a patent. Although the company can work with customers to deploy any device from the sensor to the gateway, its core IP is still in its algorithm.
The algorithm starts the entire process by breaking down the input data into multiple stream events or immediate parameters. These can be alarms, trips, or state transitions, then subdivided based on the frequency and structure of the data. Next, Cerebra applied the relevant meta-models of the electrical, hydraulic or thermal subsystems, the various failure modes of these subsystems, and the main reasons for these failure modes. Each metamodel consists of 180 data elements spanning 65 entities.
"We converted all the data lakes into the Cerebra metamodel. Once the models were properly formed, engineers could get a complete view of the Cerebra platform," said Jose.
Inputs to the model can include sensors, environmental conditions, maintenance history, historical events, and even relationships between assets and neighboring assets. Even if it is not connected, a hot or electrical noise may affect the nearby circuit board or system. Few systems can work in completely isolated environments.
High sensitivity
The deep learning algorithm continuously decomposes the much-noise data waveform into elements that Flutura calls “feature vectors†and analyzes them. Jose said: “We have about 220 waveform feature vectors. They uniquely describe a waveform. There are many IPs on top of this.†“Once you start to understand these feature vectors in depth, you will find that Time series data itself, which allows you to obtain information about your assets in a better way."
The sensitivity of the algorithm to small changes in waveforms, and its ability to correlate these disturbances with fault sequences that occurred in the past, really made Cerebra unique.
Reduce costs and new revenue models
As data is accurately collected and analyzed, more advantages will follow. For example, the user of the data may be anyone from the field engineer to the remote maintenance team. If routine maintenance may require the deployment of trucks or helicopters, the analysis will determine if it is really needed, thereby saving on overhead costs.
At a higher level, data can be used to improve quality and improve processes, further reducing costs. For equipment suppliers, service contracts for the sale or rental of equipment can be modified based on geographical location and environmental conditions: North Dakota is different from Texas or Venezuela. In addition, as far as equipment responsibility is concerned, if a device fails on site, the supplier can determine if the conditions of use at that time are not covered by the warranty.
Furthermore, predictive analytics has many benefits because it not only reduces downtime, but also provides equipment vendors and OEMs with the opportunity to develop a predictive as a service, a monitoring as a service, or a diagnostic as a service business model. Come for extra recurring income.
"We call it the digital umbilical cord," said Jose. "This is what we have done."
Many people will also be able to generate a new recurrent income stream as the "Holy Grail".
Make your device smart
Cerebra itself is hardware-independent, but Flutura has helped customers implement in the Dell Edge Gateway 5000 Series, which uses the Intel Atom processor.
Although Cerebra is hardware-independent and resides in the software stack of the edge device, it does rely on deep learning algorithms to detect the most important function that caused the failure. Although the algorithm has not yet been benchmarked for various processing architectures, Jose pointed out that performance can be improved by using Intel processors. Fortunately, given that "for most of the gateways we use, I would say that 95% of gateways have embedded Intel processors," said Rick Harlow, executive vice president and head of Americas at Flutura.
Make the most of engineer's point of view
Engineers responsible for monitoring, diagnosing and repairing large equipment from the factory to the oil rig need to view the information in a meaningful way. Flutura's Cerebra platform uses deep learning algorithms to present data from an engineer's perspective.
As a result, equipment owners can reduce downtime and improve processes, while equipment vendors can obtain new recurring revenue streams, including "predictive as a service" and "diagnostic as a service."
hdmi to usb adapter ,hdmi type c ,micro hdmi ,usb to hdmi adapter ,displayport to hdmi adapter ,usb to ethernet adapter
Pogo Technology International Ltd , https://www.wisesir.net