Today I’ll tell you about the concept of the lower concentration limit of flame propagation, an attempt to make a vibration sensor on an accelerometer, show photos from tests, and a little bit about the little things... At the end of the article there will be a video on the topic, with a demonstration of the operation of a third-party device and even a small explosion of an ATM. In the end, the device turned out, but it was too late: competitors working professionally in this area turned out to be faster and cheaper, and the trend of explosions seems to be declining.
A little about explosive gases
Some users, as it seemed to me from the comments, have a misunderstanding of the physics of gases. I will dwell in a little more detail on why a gas sensor is installed in an ATM security device.NCPRP and VKPRP
Lower (upper) concentration limit of flame propagation (NKPRP and VKPRP) - the minimum (maximum) concentration of a flammable substance (gas, flammable liquid vapor) in a homogeneous mixture with an oxidizer (air, oxygen, etc.) at which it is possible for a flame to spread through the mixture to any distance from the ignition source (open external flame, spark discharge).
If the concentration of a flammable substance in a mixture is less than the lower limit of flame propagation, such a mixture cannot burn and explode, since the heat released near the ignition source is not enough to heat the mixture to the ignition temperature. If the concentration of the flammable substance in the mixture is between the lower and upper limits of flame propagation, the ignited mixture ignites and burns both near the ignition source and when it is removed. This mixture is explosive. The wider the range of flame propagation limits (also called flammability limits and explosive limits) and the lower the lower limit, the more explosive the gas is. If the concentration of a flammable substance in the mixture exceeds the upper limit of flame propagation, then the amount of oxidizer in the mixture is insufficient for complete combustion of the combustible substance.
The values of LKPRP and VKPRP are influenced by the following factors:
— Properties of the reacting substances;
— Pressure (usually an increase in pressure does not affect the NCPRP, but the VCPRP can increase significantly);
— Temperature (increasing temperature expands the CPRP due to increasing activation energy);
- Non-flammable additives - phlegmatizers;
The addition of a phlegmatizer to the mixture lowers the value of the VCPRP almost proportionally to its concentration up to the phlegmatization point, where the upper and lower limits coincide. At the same time, the NPRRP increases slightly.
The shorter abbreviations "NKPR" and "VKPR" are usually pronounced.
In practice, the NKPR is of greater importance, because as long as the concentration has not reached this level, an explosion will not occur, and most safety systems for industrial facilities operate on the basis of this. For those interested, here is a reference table of gases and vapors where LEL values are indicated for most gases.
Alarms for sub-explosive concentrations are usually set to a certain threshold from 10 to 20% LEL, which allows early detection of an explosive gas leak. Thus, a signaling device set to 10% LEL inside the ATM safe will give a small margin of time for the security team to respond. Of course, the time there will be measured in tens of seconds, maybe minutes, but it’s still better than an alarm going off after the fact, when an explosion has already occurred. The signal from this sensor can be combined with other types of protection: from a loud siren, which can simply scare away robbers, to a cylinder with a phlegmatizer, which will make an explosion impossible.
This also leads to one more point: why you can’t do a periodic spark, as some users suggested. As long as the gas concentration inside the ATM volume is less than 100% LEL, igniting the mixture will lead to nothing, and as soon as the concentration exceeds this threshold, an explosion will occur. In this case, internal damage is inevitable. And who can guarantee that this energy will not be enough to open the safe door? In short, you can't do that.
Vibration sensor
Of course, we had hopes of killing two birds with one stone: tracking both tilt and vibration. Compel specialists kindly helped us with the choice of accelerometer, offering the following options:Anatoly, our engineer recommends choosing a new aclerometer. LIS331DLH is the oldest and most expensive.
We chose LIS2DH12 and immediately ordered debugging boards for experiments. The debugging kit included very sophisticated software, in which it was possible to observe graphs for each of the axes in real time, perform a Fourier transform, etc.
Here's a video from ST with this kit:
With this set, I tried to understand whether the accelerometer could sufficiently sense the vibration from the impact of the tool. Just like this, they attached it to the wall of the test ATM safe and drilled/sawed it.
Such an accelerometer is capable of producing up to 5.3 kilosamples per second, respectively, the Nyquist frequency will be 2.7 kHz. Purely theoretically, this would be enough to cover the range of vibration created by the instrument, but in reality we received too little sensitivity.
Nyquist frequency
Nyquist frequency - in digital signal processing, a frequency equal to half the sampling frequency. Named after Harry Nyquist.
From Kotelnikov’s theorem it follows that when sampling an analog signal, there will be no loss of information only if (spectral density) the highest frequency of the useful signal is equal to half or less than the sampling frequency (in the English literature, the term Nyquist frequency is used to denote half the sampling frequency). Otherwise, when restoring an analog signal, there will be an overlap of spectral “tails” (frequency substitution, frequency masking, aliasing), and the shape of the restored signal will be distorted. If the signal spectrum has no components above the Nyquist frequency, then the signal can (theoretically) be sampled and then reconstructed without distortion. In fact, the “digitization” of a signal (conversion of an analog signal into a digital one) is associated with the quantization of samples - each sample is written in the form of a digital code of finite bit depth, as a result of which quantization (rounding) errors are added to the samples, under certain conditions considered as “quantization noise”.
Real signals of finite duration always have an infinitely wide spectrum, which decreases more or less rapidly with increasing frequency. Therefore, signal sampling always leads to information loss (distortion of the signal shape during sampling and reconstruction), no matter how high the sampling frequency is. At the chosen sampling rate, distortion can be reduced by suppressing spectral components of the analog signal (pre-sampling) above the Nyquist frequency, which requires a very high-order anti-aliasing filter to avoid aliasing of tails. The practical implementation of such a filter is very complicated, since the amplitude-frequency characteristics of the filters are not rectangular, but smooth, and a certain transition frequency band is formed between the passband and the suppression band. Therefore, the sampling frequency is chosen with a margin
And we were unable to adequately calculate the inclination angle at such a sampling frequency, the noise was too great. Therefore, we reduced the operating frequency of the accelerometer to 50 Hz and can easily obtain the tilt angle from it, and the role of a vibration sensor is performed by a massive metal plate with a piezo block.
What I would like to mention regarding the features of the vibration sensor operation algorithm... Here is an excerpt from the instruction manual for one of the detectors on the market. It follows that you can drill/knock, but not more often than once every 10 seconds, then the detector will not work - this is protection against single impacts. True, I won’t take responsibility for the strength of this single blow and the intensity of drilling; perhaps there is a certain threshold set, above which the detector will still work.
Tests
For testing, we came to the bank's security service, they rolled out an old ATM like this (it's in the photo in the header, only closed):
The trunk of my car at that time looked like this:
They were attached to the wall of the safe, powered by a battery:
Like this In this way, we debugged and carried out the tests we needed: we supplied gas, drilled, smoked, tilted, etc. Unfortunately, or maybe fortunately, the security service did not allow us to blow up the ATM.
For trial operation, we were allocated a combat ATM in one of the city’s branches next to the tram tracks. It was supposed to evaluate whether trams passing by would lead to false alarms: the rails in our city are dead, and most of the trams are old - the earth shakes when they approach. But overall, there were no complaints during the month of operation of the device.
I guess I'll end my story here. The device turned out to be good, but it is unlikely that it will be possible to sell it due to certain circumstances and competition.
