This week Melinda talked with Dr. Beatrice Alex who is in computational linguistics at the University of Edinburgh. I learned a lot about when and how to include subject matter experts in the NLP and semantic object identification pipeline. We can take some of the lessons she learned in their digital humanities project into our Siri for Maintenance project. Beatrice’s homepage is http://homepages.inf.ed.ac.uk/balex/
The traditional methods for high frequency data analysis involve many piezoelectric sensors, signal amplifiers and spectrum analysers the size of a modern day desktop computer. However, with the arrival of low cost and easy to use microcontrollers and Micro Electromechanical Systems (MEMS), perhaps it’s time to reconsider some of the traditional methods. This blog post aims to outline some of the challenges, limitations and success encountered whilst trying to use a combination of microcontroller and MEMS devices for condition monitoring.
- Arduino MEGA 2560: . The large user community along with extensive libraries available make this an ideal choice for experimentation.
- ADXL345: The ADXL345 is a small, thin, low power, 3-axis accelerometer with a measurement range of upto ±16 g.
Thanks to the advances in rapid prototyping, it is incredibly easy to produce PCB’s in low quantities whilst keeping cost to a minimum. We had designed and manufactured our own PCBs to support this project. The PCB houses the accelerometer, a temperature sensor, a voltage and current sensor and a various other electronics. This test setup cost us <$100.
Challenge #1: Data storage:
The Arduino lacks an onboard flash storage. Therefore it requires an external storage space to store the data. An SD card shield is the simplest solution to overcome this problem. This method however, has one big disadvantage and that is the speed at which data can be written to an SD Card. Every time the Arduino has new data to be written to the SD Card, a file in the SD Card has to be opened, the data written and then closed. Failure to close it at the end of each data entry may result in corrupt data. The process of writing to an SD Card is slow and is subsequently one of the biggest limitations to recording high frequency data.
Solution: Buffer memory
The Arduino has 512kb of Random Access Memory (RAM) which has a much higher read/write speed than any external flash storage. So by writing to this memory, then transferring it in chunks to an external flash storage at once, we are able to record data at much higher frequencies. However, the measly 512kb RAM posses severe limitations to the number of data points that can be obtained. We are limited to around 500 data points before the the RAM fills up and requires to be transferred to the SD Card. The Arduino is also incapable of collecting any new data whilst this transfer is in progress. This means that the vibration data is not continuous and can only be obtained in small chunks. There are methods to increase the RAM of the Arduino using external RAM modules, which would enable us to increase our sample size significantly. But that is a different challenge on it’s own, and one we will hopefully tackle in the near future.
Challenge #2: Obtaining consistent data
This issue is most relevant for vibrational data.
The ADXL345 supports a maximum Output Data Rate (ODR) of 3200hz on SPI and 1600hz on I2c. Our test rig is currently configured to use I2C. To understand this challenge, some knowledge of the Arduino’s operation is required: The Arduino continuously carries out a set of instructions in a conditional loop. It can be configured to request data from it’s sensors each time the loop is executed. The time taken to complete the instructions inside the loop is very inconsistent and therefore, the time interval between each data point can vary. This inconsistency can severely affect some time sensitive data such as vibrational data. Without a fixed ODR, the vibration data is essentially useless.
Solution: FIFO Buffer
Luckily, one of the ADXL345 chip’s distinguishing features, is the inclusion of a First In First Out (FIFO) buffer. The chip continuously collects data at a fixed rate, stores it in it’s FIFO buffer, and triggers a watermark when the FIFO buffer is full. The Arduino can be configured to continuously scan for this watermark and transfer the data from the FIFO buffer onto it’s RAM upon the watermark’s signal. This guarantees a fixed, consistent ODR. Ofcourse, this only works because the Arduino can poll the accelerometer at a faster rate than the accelerometer is collecting data.
Challenge #3: Accuracy
Coming soon: We are testing this and will update shortly.
System Health Lab, UWA
So recently we were doing some data collection with an Arduino mega and an accelerometer, specifically the ADXL345. Now the datasheet on the ADXL345 stated that the maximum sampling frequency is 3200 Hz but we found that our data points we only coming through at about 900 Hz. After some digging around the web, we found a potential cause.
Although the Arduino should be able to handle up to 400 kHz I2C communication, it is by default limited to 100 kHz in a particular library header file. Now 100 kHZ should be more than enough to sample a device which is supposed to be capable of 3200 Hz sampling but it’s not as straightforward as that. We think that although 100 kHz is the I2C bus speed, there is still going to be some time initiating communication. Depending on the library you choose to use. there could be several back and forths between the Arduino and the ADXL345 before any data is actually exchanged.
We found however that by making the change to 400 kHz, we were able to obtain data at 2600 Hz which while not maxed out, is still a significant improvement! Perhaps the remainder is due to the aforementioned communication overhead. Here’s what we did to obtain the increased bus speed as laid out at this link.
Locate the twi.h header file at this location: C:\Program Files (x86)\Arduino\hardware\arduino\avr\libraries\Wire\utility.
Once you’ve located the file, note the highlighted section. This is the default value provided when you download the Arduino software. Change the 100000L to 400000L. It’s a simple as the picture below!
And that’s pretty much it!
You’ll need to let Arduino recreate the associated object files but this is simply a matter of restarting the Arduino IDE. You should notice an increase in your I2C communication! Mind you that you’ll have to be talking to a device capable of making use of this increased speed (like we had with the ADXL345).
I won’t claim to know all the maths behind this so if you know more, please share your knowledge with us in the comments section! Alternatively if you found this a useful link, leave that in the comments too!
System Health Lab, UWA
Once we knew what we wanted to do, all that was left was to do it! This was not a quick process however. We discovered that knowing what we wanted to do and knowing exactly how to go about doing it were quite different things. There was a lot of Googling involved. Thankfully, most of our design tools were popular and well documented, specifically Solid Works for the mechanical aspect and Arduino for the electronic and programming side of things. We also drew heavily on our previous experience, pulling ideas and skills from seemingly unrelated previous projects. During the early stages, there was not much programming as we focused more on building a basic test rig, even if it lacked sophisticated sensors and data communication. As such, Solid Works was used a lot. We would design the different components separately and then use CAD to see how it fit it within the larger assembly. There were many iterations to get it all right but eventually we completed it. Once we were satisfied with our CAD drawings, we took it to the workshop and had it produced.
After listing down our design considerations, we could move on to more specific designs. Concept level design decisions were made using decision trees. Here’s a specific example. We knew that as a truck, our test rig would have to have some form of driving. So we brainstormed the various options available: electric motor, heat engine, etc. We then qualitatively assessed each of these options again the design considerations we had. Based on this, a decision was made that impacted the next step in the decision tree. In this case, we decided to go with the electric motor option and then had to consider how to power it: battery, charged track, etc. We would go back to the assessment stage and keep carrying out this process till we reached the end of the tree, a very specific decision that could not really be debated on. We kept records of these trees in case we ever needed to back track at a later stage.
First and foremost, we had to list down the design considerations for this project. There were a few key areas that helped us narrow down our focus:
- Functionality: The test rig had to have controlled failures which meant that parts that were not supposed to fail had to be over-engineered! Even if this meant choosing a slightly more expensive motor driver with excessive current sourcing capabilities, there was no competition if the closest rival could fail under normal test conditions.
- Simplicity: We wanted the design to be simple enough so that even a 1st or 2nd year undergraduate student could understand, even contribute to, the project. To aid this, we also aimed to minimised the number of manufacturing processes undertaken so that a ‘standard’ student would be easily able to pick up what was going on. In keeping this simple, it helps to have fewer components. Therefore we aimed to have multi purpose components where possible. An example of this was using the Arduino platform as it is incredibly diverse and able to carry out many of the necessary functions. Finally, we also tried to keep components reasonably priced.
- Awesomeness: While this might seem like an odd criteria, we recognised that what we were doing was useful and incredibly cool as well! We wanted to impress anyone who came into contact with this project. This is one of the reasons why we chose to model the truck after actual mining trucks, bright yellow and all!
- Upgrades: The final factor we kept in mind was future expansions on the test rig. We wanted a rig that future team members could easily add to or change as necessary.
Preliminary tests on the operating range and control of the environmental variables within the chambers has been conducted. Results are very positive, with humidity generation and control exceeding expectations, and temperature extremes being easily attainable with no failures. Unfortunately UV output is less than expected.
Humidity Control Tests:
Humidity control is required for both chambers. For the UV chamber preliminary testing shows that the humidification system can raise the humidity level in the chamber to 80% in under 5 minutes from an ambient of ~35%RH.
With the testing parameters used the Humidity system struggled to achieve 90%RH in the given time. The dehumidifier is effective, but slow. It is effective at maintaining a low ambient humidity, but not rapidly lowering the humidity level.
We can see that control is excellent. Achieving tolerances below +-1%RH. Additionally, we can see that the humidifier has minimal impact on the temperature of the system.
For the oven we can see that control is also excellent. Multiple temperatures are yet to be tested.
Temperature Control Tests:
Temperature is being controlled by a simple Bang-Bang control algorithm. A PID algorithm has also been implemented and tested, but results are still poor due to additional parameter tuning required.
The graph below illustrates temperature control within the oven with 15 minute set point intervals and no insulation. Insulation has been added since this preliminary test. We can see we easily achieve the required temperature of 120C, and can control temperature within +-2C. With insulation and a tuned PID it is expected that these results will improve. However, this test confirms the core functionality of the temperature system.
Ultraviolet (UV) System:
The UV system is designed to output in the UV-B spectrum, to maximise degradation. Initial tests have shown that UV-B levels within the chamber are approximately equal to that of Peak Sunshine on a Hot Summers day in Perth. This corresponds to an output of 3mW/cm^2.
Output is fully controllable between this maximum value and 20% of this maximum.
We had hoped for an output of approximately 5 times this.
Above is the Master Wiring Diagram. This schematic details all the electrical connections required for the ALT project. Additionally, it labels all the micro-controller pins required. The Arduino Mega is used to process all I/O functions, while the Arduino Yun manages data-logging functionality.
Sub-systems have been kept together and connections spaced out to aid debugging. Certain time critical functions are attached to interrupt pins, such as the Door switches and safety resets.
Finally, as a fail-safe, certain sensors report to both the Arduino Yun and Mega. This allows either micro-controller to reset the other in case of an error or lock up. Each processor also has control of safety critical functions and can cut power if required, minimizing risk.