Why Predictive Maintenance Is Fundamental in Industry 4.0

{tabbedtable} Tab Label	Tab Content
About	Cost, safety, availability and reliability are the main reasons why key industrial players are investing in predictive maintenance. Predictive maintenance allows factories to monitor the condition of in-service equipment by measuring key parameters like vibration, temperature, pressure and current. Such monitoring requires connected smart sensors featuring a high-speed signal chain, powerful processing, and wired and/or wireless connectivity. The current trend is leaning toward wireless connectivity because it is possible to cover remote locations and does not add extra wiring costs. Wireless connectivity converts existing equipment into smart equipment (rather than replacing it), enabling users to plan maintenance based on actual needs and increasing productivity and efficiency. That is the ultimate goal of Industry 4.0.
Presenter	Giovanni Campanella, Systems Engineer Giovanni Campanella is a systems engineer at Texas Instruments, where he is responsible for defining and developing reference design solutions for the industrial segment. He earned his bachelor's degree in electronic and telecommunication engineering at the University of Bologna and his master's degree in electronic engineering at the Polytechnic of Turin in Italy. He is an expert in sensors and analog signal chain.
Webinar Recording	Click here to view the recording

{tabbedtable} Tab Label

Tab Content

About

Cost, safety, availability and reliability are the main reasons why key industrial players are investing in predictive maintenance. Predictive maintenance allows factories to monitor the condition of in-service equipment by measuring key parameters like vibration, temperature, pressure and current. Such monitoring requires connected smart sensors featuring a high-speed signal chain, powerful processing, and wired and/or wireless connectivity. The current trend is leaning toward wireless connectivity because it is possible to cover remote locations and does not add extra wiring costs. Wireless connectivity converts existing equipment into smart equipment (rather than replacing it), enabling users to plan maintenance based on actual needs and increasing productivity and efficiency. That is the ultimate goal of Industry 4.0.

Presenter

Giovanni Campanella, Systems Engineer

Giovanni Campanella is a systems engineer at Texas Instruments, where he is responsible for defining and developing reference design solutions for the industrial segment. He earned his bachelor's degree in electronic and telecommunication engineering at the University of Bologna and his master's degree in electronic engineering at the Polytechnic of Turin in Italy. He is an expert in sensors and analog signal chain.

Webinar Recording

Click here to view the recording

Attachments:

Why predictive maintenance is fundamental in Industry 4.0.pptx

Hi all,

There were some questions that were not addressed during the webinar. Below the answer:

"Are AlexNet, ResNet supported? Can I train and deploy my own custom models?"

The networks themselves would need to be slightly modified in order to be compliant with our input tool in order to get them to run on the AM57x processor. Note that these networks are not designed for an embedded processor and are more used for benchmarking in the industry.

Yes you can train and deploy your own custom models. We provide an embedded friendly network Caffe-Jacinto to help with this. For custom models, there are restrictions on the layers and parameters supported, but if you stay within these restrictions they can deploy on our AM57x devices.

"What are the best ways to reduce false positives and false negatives?"

It is a mix of having the good neural network design and having enough training data / not biased. More training data will reduce the false positives and false negatives.

Thanks

Giovanni

Webinar slides are attached.

This event is open for registration.

Top Comments

rscasny over 5 years ago +2

This event is open for registration.

gio89 over 5 years ago

Hi all,
There were some questions that were not addressed during the webinar. Below the answer:
"Are AlexNet, ResNet supported? Can I train and deploy my own custom models?"
The networks themselves would need to be slightly modified in order to be compliant with our input tool in order to get them to run on the AM57x processor. Note that these networks are not designed for an embedded processor and are more used for benchmarking in the industry.
Yes you can train and deploy your own custom models. We provide an embedded friendly network Caffe-Jacinto to help with this. For custom models, there are restrictions on the layers and parameters supported, but if you stay within these restrictions they can deploy on our AM57x devices.
"What are the best ways to reduce false positives and false negatives?"
It is a mix of having the good neural network design and having enough training data / not biased. More training data will reduce the false positives and false negatives.
Thanks
Giovanni
- Cancel
- Vote Up 0 Vote Down
- Sign in to reply
- More
- Cancel
rscasny over 5 years ago

Webinar slides are attached.
- Cancel
- Vote Up 0 Vote Down
- Sign in to reply
- More
- Cancel
rscasny over 5 years ago

This event is open for registration.
- Cancel
- Vote Up +2 Vote Down
- Sign in to reply
- More
- Cancel

Why Predictive Maintenance Is Fundamental in Industry 4.0

When

Where

Event Type

Top Comments