Thermal Detection System

[shockie]

Attached is a screenshot of my first iteration of what I am calling a thermal Detection System. I have a small weather station perched atop a 10m tall thermal pole. The weather station is an esp32 mcu that is sending the sensor data back to a raspberry pi4b via 2.4ghz wifi, to an influxdb database and the data visualization is done thru Grafana, both running on the rpi4b under Raspberry Pi OS, a Debian linux
distro.

anybody at the field can access the web page by simply connecting via wifi
to the Rpi4b which is in AP mode So a tablet,laptop,phone etc can all see the same data in realtime using their browser on their device.

If there was mutiple thermal pole weather stations, the data from each could be displayed as a tab in the browser or I could consolidate the data and show it combined on one browser tab.

still got a long way to go, but I'm getting there.

 

[cls]

what is a thermal pole?

 

[shockie]

what is a thermal pole?

It's a pole that has a mylar streamer attached to it. Mine has a small weather station attached to the top.

I'm seriously considering placing a camera at a 90 degree right angle so I can watch the streamer. the camera would stream the video via 2.4ghz wifi back to the rpi4b in a browser tab.

This way I would have both visual cues and actual realtime data.

what is a thermal pole?

 

[David_Stack]

Many years ago as a fund-raiser for the US F3J (Radio Control Thermal Duration sailplanes) Team heading to the World Championships Joe Wurts (amazing sailplane pilot) put together a video on tips to locating thermals. He used three poles on the flying field with mylar tape, and could very effectively 'triangulate' where thermals were likely to be found based on observing the mylar 'telltales' on each pole.

He routinely hand-tossed his 3M span sailplane into thermal after thermal, gaining altitude, coming back to land, then doing it again...

 

[cls]

oh, now I get it! thanks

 

[shockie]

Here's an update to my Thermal Detection System. Still calibrating the wind direction. Still much work to do. ESP32 > Raspberry Pi (via wifi) > Influxdb > Grafana

 

[Buckeye]

How is the weather station data used to detect thermals?

 

[shockie]

well the first thing you are looking for is a quick rise in temperature of say 2 to X degrees in a short time period. This signifies that there is warmer air being sensed by the temp sensor. Another is the interaction between a rising temp is humidity. The relation between humidity and temperature is inversely proportional. If temperature increases, it will reduce relative humidity; thus, the air will become drier. When the temperature decreases, the air will become wetter; therefore, the relative humidity will increase.

so you have 2 indicators : if temp rises will fast in a small time frame AND humidity decreases then there's the real possibility of a thermal being detected.

Other indicators include wind speed and wind direction.

Normally, a thermal pole/weather station is situated upwind of your launch site. Say 100ft or so. When a thermal is detected by metrics such as above along with seeing what the thermal streamer is showing, if there's a slight breeze of 2-6 mph and all of a sudden its calm. And the wind vane sensor goes from the wind direction the wind is blowing to suddenly differing directions , all of these put together denote the real possibility of a thermal.

Traditionally the FAI Competition people have used Krestrel weather Meters which is basically a small handheld anemometer. And while it has a wireless link, its Bluetooth BLEwith a line of sight of maybe 100 ft. So they station a thermal spotter upwind with one and he relays back what he perceives to be a thermal.

Kevin Kuzcek developed another handheld temperature sniffer device .

The major drawback to hand held devices in my opinion, as it requires a human and a humans experience to correctly detect or sense a thermal.

My approach is to place a thermal pole with a small weather station on top along with the thermal streamer. the ESP32 has 802.11 b/g/n wifi so it can collect the sensor date and transmit it via 2.4ghz wifi to a raspberry pi 4 or 400. (line of sight is approx 1000 ft).The Raspberry pi has installed on it a time series database called Influxdb and a data visualiser called Grafana that takes the data in the Influxdb and displays the sensor data in near real-time

Additionally if you had multiple Thermal Detection systems out in the field, the data can be displayed [er Thermal detection pole or you can show an aggregate view from all 3 on one page; one page being a web browser web page. So anybody that has a phone,laptop,pc,etc at a competition can attach to the raspberry pi which is configured as an AP hotspot.

so in essence you get a networked thermal detection system. You can add as many thermal detection units as you see fit.

The late (grreat) George Gassaway I believe pioneered using bubble machines in conjunction with a thermal pole, to help visualise the thermals. You can also put out a few score of 4 ft high fiberglass poles with streamers on them upwind to help you spot thermals.

I pan to create a pcb and a "sled" to put the electronic components in atop my 10m thermal pole.

Later I plan to use a raspberry pi with a pi camera at a right angle to the thermal pole so I can visually see
the thermal streamer in real-time. The camera will be able to stream video to my launch site. So If the sensors sense a thermal I can watch the thermal pole in real-time to see how its behaving.

 

[shockie]

Here's an example of what a thermal "signiture' might look like.

Notice the rise in temp and the decline of humidity?

Also notice to the right, just the opposite. humidity rising and temperature declining.

Showing the relationship between temp and humidity.

 

[astronwolf]

Here's an example of what a thermal "signiture' might look like.

Notice the rise in temp and the decline of humidity?

Also notice to the right, just the opposite. humidity rising and temperature declining.

Showing the relationship between temp and humidity.

You are just showing micro shifts in temperature and humidity over four seconds. I'm skeptical.

 

[shockie]

You are just showing micro shifts in temperature and humidity over four seconds. I'm skeptical.

Of course you are. That was an example of what a thermal signature might look like. I used a hair dryer blowing warm air near the sensor to simulate a thermal.

The point is, a quick rise in temp with a corresponding decline in humidity is a good indicator of a thermal.

Also the time frame is approximately 16:59:30 to almost 17:01:00; that's not 4 seconds, it's about 90 seconds.

So your skepticion is based on an incorrect reading of the time scale. The time scale is in 30 second increments.

 

[astronwolf]

So your skepticion is based on an incorrect reading of the time scale. The time scale is in 30 second increments.

My skepticism is based more on watching people consistently catch thermals without using electronic gizmos and watching people completely miss thermals while squinting at those kinds of devices. That doesn't mean that I think those devices you are are working on aren't going to work. But I do believe that ordinary human senses do an amazingly good job at spotting thermals. Hence, I'm skeptical and biased by what I've witnessed people do.

Kind of like this guy:

 

[kjhambrick]

+1 for astronwolf's video

+1 for shockie's weather station because I always record Temperature and Barometric Pressure at the Launch Site so I can post-process Flight Computer Data.

It would be nice if there was a wireless weather station at the Pads so I didn't have to use my old-timey wall-mount

-- kjh

 

[shockie]

Well one of the impetsus behind this project is to bring "picking good air" to the masses.

Every FAI competitor that I've ever spoken to about thermals insist that you simply cannot win a duration event no matter how high tech your model or launcher may be. You got to be able to pick good air is the bottom line.

I agree that human senses do a really good job of being able to pick good air; the downside is that it takes experiences over time to develop those senses.

But this is the 21st century and using human senses and experience is so 20th century.

Another impetsus was to provide competitors with little to no experience picking good air to perhaps learn quicker and easier how to recognize good air with the aid of technology.

Let's call this human assistive technology.

I have spoken at great length with a local Meterologist peppering him with questions about Thermal formation and detection methodologies.

I then used my questions and his answers as questions with ChatGPT.
The results were very interesting and enlightening .


This hardware/software combo is just one piece of the puzzle.

Part 2 will be a pi camera recording the thermal pole streamer at a right angle.

This will show when the thermal streamer detects a thermal.

Each video frame had a timestamp.
I can then correlate the thermal streamer with the the environmental data timestamps.

Part 3 will be develop a machine learning model and train it for objection detection of the thermal streamer and then object tracking of the thermal streamer.

Machine learning is the basis of AI.
You create a ML model with data which can be actual numerical data and or video data where each frame of the video is "read" by a python algorithm for the object it is detecting and tracking.

So basically I'm trying to create an AI to detect thermals.

I had hoped to have this all done by now, but with covid, electronics supply change issues, it's been difficult if not right down impossible to get the required hardware at reasonable prices. I'm not paying a 150 to 200% markup. The availability and prices are due to come down later this summer so I expect to be able to start interfacing the parts into a whole.

 

[shockie]

Here's a look at version 2 of my electronics package to be. On the left is an ESP32 DOIT Devkit C V1 that has 3 sensors attached: a bme280, an Ah49E Hall effect sensor, and an AS5600 Magnetic rotary encoder. The bme280 of course measures temp, humidity, and barometric pressure. The AH49E hall effect sensor has a magnet that rotates around it allowing me to measure wind speed. The AS5600 magnetic rotary encoder allows me to measure wind direction by use of a rotating diametric magnet above the sensor. This is a unique magnet in that one half is S pole and the other half is N pole.

In Ver 1 I was using magnetic reed switches ( and magnets) to determine both wind speed and direction and they limited how many cardinal points on a compass that I could measure. put another way, much lower resolution. with the HES and MRE I can now measure speed and direction more precisely. The magnetic reed switches was limiting me to 8 cardinal points, now I can do 16.

The new ESP-32-ISO board allows me to power the entire unit with Ethernet POE, Power over Ethernet, as the entire unit will be 10m up in the air. It also has a battery attachment where i can attach a backup lipo battery. So there's some power redundancy. The ESP-32-ISO is connected to a POE Injector (or POE switch) which is plugged into a Bluetti EB3A Battery onsite. You might ask whu not use a 10m+ length usb cable for power ? Well, first they don't make usb cables that long except as a custom cable and it requires electronics built into the cable, so they are way too pricey. I'm not using the ethernet cable for data, just power. The ESP32 transmits it's sensor data to a raspberry pi 4b base station via 2.4GHz wireless.

I will be adding an AD1115 ADC because the onboard ADC(s) have a linearity problem discerning the difference between say 3.2 to 3.3 volts on the upper end.

I've attached a small .mp4 video you might find interesting. be sure to turn on the sound .