Method and system of waveform acquisition and management on IoT and IIOT devices and systems aimed at a spectrum analysis

Abstract

A method of waveform acquisition and management on IOT and IIOT devices and systems aimed at a spectrum analysis comprises the steps of sampling according to regular time intervals one or more signals from a physical emitter by means of one or more data acquisition and processing devices and obtaining a sampling vector; processing the sampling vector data by calculating the RMS of the sampled data; processing the sampling vector data by performing a Fourier Transform calculation of the sampling vector and obtaining an FFT vector representing a spectrum of the one or more sampled signals; performing a frequency resolution reduction of the FFT vector data; performing fractional recoding of the data obtained from the previous step to reduce the number of data needed to transfer the useful information; performing bit compression coding of the data from the spectrum reduced in the previous step; and transmitting the data thus processed by means of a radio transmission system employing a long-range, low-power transmission mode.

Description

[0001] Title

[0002] Method and system of waveform acquisition and management on IOT and HOT devices and systems aimed at a spectrum analysis.

[0003] Description

[0004] The present invention refers to the field of monitoring and control systems. More specifically, the invention concerns a method and system of waveform acquisition and management on IOT and HOT devices and systems aimed at a spectrum analysis.

[0005] IOT and HOT systems, “Internet of Things" and “Industrial Internet of Things", respectively, consist of appropriately extended sets of autonomous and interconnected devices. Autonomous, because each device in the system is capable of performing specific functions. Interconnected because each device is connected, directly or indirectly, to one, some or all of the devices in the same system.

[0006] IOT and HOT systems are also used to build and control sensor systems aimed at acquiring and analysing, even automatically, conditions that, for various reasons, should be monitored and controlled.

[0007] However, the use of IOT and HOT systems in this context has some drawbacks due to two main needs: making IOT and HOT devices characterised by very low power consumption, with the possibility, where necessary, of powering them with small solar panels or batteries; and covering areas as large as several kilometres or tens of kilometres, using communication networks internal to the systems (i.e. not third-party communication networks such as 4G or the like). For these reasons, most known IOT and HOT systems use the Internet and Internet-specific protocols only in some elements of their architectures.

[0008] For instance, several alternative communication vectors to those traditionally used for the Internet have been developed that are capable of establishing medium- and long-range wireless digital communication networks. The best known is probably Lo.Ra., but there are also zigBee, SigFox, Esp Now and others. These are communication vectors that can reach up to a range of several kilometres and are characterised by very low power consumption and frequencies that are normally lower than those of the widespread WiFi network.

[0009] These solutions also present some critical issues. For instance, the characteristic of these communication technologies is that they can send messages containing a limited number of bytes and are much slower than home WiFi networks.

[0010] In addition to the above, in some cases, IOT or HOT devices are used to acquire and analyse signals that are in the form of waves, and, in many cases, the actual usefulness of acquiring signals that are in wave form is obtained by subsequently analysing their spectrum. These may include, but are not limited to, sounds, vibrations, radio emissions and the like.

[0011] An example of this type of application is predictive maintenance. In this case, the common practice is to attach one or more accelerometers to an industrial machine in order to detect vibrations and analyse them. From the variation of vibrations over time and from the intensity of the vibrations themselves, it is possible to predict failures, anticipating sudden production stoppages and side effects, with considerable advantages in terms of operating costs, effectiveness and efficiency of the systems. However, while the intensity of vibrations can be useful for identifying the presence of a problem in the vicinity of a sensor, in order to investigate its causes, it is necessary to appropriately analyse the spectrum of the vibration waveform itself in order to assess its prevailing frequencies.

[0012] Precisely in the case of predictive maintenance, two ways of detecting vibrations are known. Portable devices or permanent devices can be used. In the second case, the devices, usually IOT devices, are placed and left for a long time in the vicinity of the signal source, and are able to take the necessary measurements and send them to a central server for analysis, notifications following the identification of any anomalies, and more.

[0013] In general, making permanent sensor networks to be applied in contexts that need to monitor measurements in the form of waves is far from a simple task.

[0014] Keeping track of a measurement that is expressed as a single number that can be acquired at regular intervals of a few seconds, as is usually the case for temperatures, for example, results in a very small amount of data, and there are no problems in the use of vectors such as Lo.Ra. and the like because each piece of data can correspond to a message.

[0015] On the other hand, when it is necessary to acquire and analyse data characterised by variations over time with high frequencies of tens, hundreds, thousands or hundreds of thousands of Hz, it is a different matter. If we were to acquire data from 100 3-axis sensors with a sampling rate of 3200 Hz, we would need to acquire and transfer, for each sensor, 9600 numbers per second. This would result in a data volume of approximately 154 Gb per day.

[0016] In the case of ultrasonic vibrations, i.e. with a sampling frequency of 100 Khz, the amount of data would increase enormously, so much so that the actual application would be practically infeasible or in any case too expensive for the purpose.

[0017] However, the typical IOT transmission vectors mentioned above are, by their very nature, not capable of continuous data streaming. Consequently, they cannot be used for the continuous transmission of time-varying data with high frequencies.

[0018] For this reason, a plethora of solutions have appeared on the market based on the concept of “edge computing’, i.e. the practice of distributing the burden of calculation over all the elements of the system, including the sensors and data acquisition devices.

[0019] Since IOT devices have autonomous operating capabilities, and are therefore capable of performing signal analysis in addition to simple detection, there are a number of IOT devices on the market that have the ability to calculate the onboard spectrum.

[0020] One example is patent no. IT202022000003111 , filed by Digital Media Industries S.r.L, which describes an architecture in which the Fourier transform (FFT) required for spectrum analysis is performed directly on the data acquisition and processing device, in order to reduce the amount of data to be sent to the central system, and a reduction in frequency resolution is carried out in order to better manage the data. This reduction essentially allows the transfer of Fourier coefficient data from the sampled wave. However, there is no data on the RMS nor on the physical displacement generated by the accelerations and velocity achieved by the body whose oscillation is measured.

[0021] In most known applications, there are therefore two operating modes: the data are acquired by transducers and communicated to analysis servers in full format and then analysed and stored, with the resulting cost and management problems; or the data are analysed on board the sensor and reduced to intensity measurements (RMS), or the sensor itself contains threshold values for specific frequencies, and only reports any deviations from said values.

[0022] In the first case, the amount of data to be managed for even medium complexity systems is a major problem. In addition, there is also the problem of transport, transmission, of data. It is not possible to use typical IOT or HOT communication vectors similar to Lo.Ra. to send waveform data, because these communication technologies do not allow continuous data to be sent.

[0023] In particular, it is not possible to transfer time-varying data with sampling frequencies greater than:

[0024] 1 minimum time to send a message taking into account that the actual sending time is variable and can be up to a few seconds. Clearly this is not usable for practical purposes.

[0025] Therefore, it is necessary to resort to networks or other cable-based data transfer systems, which, in some cases, may mean laying kilometres of cable, resulting in a considerable increase in costs and installation problems.

[0026] In the second case, the data item obtained loses value, which reduces the possibility of systems and practices related thereto to improve over time, either through experience or through machine learning or artificial intelligence systems in general. These modes are certainly inadequate for most research and development activities.

[0027] The ability to sample data of time-varying phenomena at “high” frequencies, representing waveforms, and to derive the spectrum so that it can be efficiently transferred using Lo.Ra.-like vectors typical of IOT and HOT, is a need that is increasingly manifested in a variety of fields.

[0028] The overall object of the present invention is to satisfy this need by maximising the level of precision of results in an extremely simple, inexpensive and particularly functional manner.

[0029] This object is achieved by the features of the invention reported in the independent claim. The dependent claims outline preferred and / or particularly advantageous aspects of the invention.

[0030] It is to be understood that elements and characteristics of one embodiment may conveniently be incorporated into other embodiments without further clarifications.

[0031] Reference will now be made in detail to the various embodiments of the invention, with particular reference to the accompanying figures, in which:

[0032] - Figure 1 is a general schematic view of the system of the present invention;

[0033] - Figure 2 illustrates a graph representing amplitude variations over time of a signal of interest;

[0034] - Figures 3a and 3b illustrate different sampling intervals of the signal in Figure 2 performed with a known sampling process; and

[0035] - Figures 4a and 4b illustrate different sampling intervals of the signal in Figure 2 performed with a sampling process according to the present invention. Each example is provided merely to illustrate the invention and is not intended as a limitation thereof. For example, the technical characteristics illustrated or described because they belong to one embodiment may be implemented in, or in association with, other embodiments in order to generate a further embodiment. It is understood that the present invention shall be comprehensive of such modifications and variants.

[0036] In this description, the term “waveforms” refers to the result of sampling data on the variation of physical magnitudes over time at frequencies above 1 Hz.

[0037] A waveform acquisition and management system on IOT and HOT devices and systems for spectrum analysis according to the present invention may comprise:

[0038] - a server, or a computer 10;

[0039] - one or more data acquisition and processing devices, hereinafter referred to as “data acquisition and processing device” 30.

[0040] Each data acquisition and processing device 30 may include in turn:

[0041] - a sensor assembly, comprising one or more transducers;

[0042] - one or more analogue-to-digital conversion (ADC) devices;

[0043] - a timer, capable of switching the data acquisition and processing device and / or the remaining components on and off;

[0044] - an electronic control and data processing unit, or processor, or other technically equivalent device; and

[0045] - a memory system.

[0046] The waveform acquisition and management system also comprises a data transmission system, comprising, for example but not limited to, one or more receiver antennas or “gateways” 20, and a data transmission device, arranged on board each data acquisition and processing device 30. The control and processing unit of each data acquisition and processing device 30 is in communication with the one or more transducers, the one or more analogue / digital conversion devices, the data transmission device, the timer and the memory system, and is configured to execute instructions stored as a programme in the memory system and to send and receive signals.

[0047] The memory system may include various types of memory, including optical memory, magnetic memory, solid-state memory and other non-volatile memories.

[0048] The data transmission device can be of the Lo.Ra., Esp-Now, Sigfox, zigBee type or other long-range, low-power transmission mode, even “mesh” mode.

[0049] The system, or each of its components, may be removable or relocatable with respect to the position of a physical emitter, i.e. a signal source. In an embodiment of the present invention, data acquisition devices are configured to acquire one or more signals of interest from one or more industrial machines 40.

[0050] The present invention also relates to a method of waveform acquisition and management.

[0051] The method comprises an initial step of acquiring one or more signals of interest by means of one or more data acquisition devices 30.

[0052] Since the data acquisition and management device 30 of the present invention is based on digital technology, the signal acquisition takes place by means of a sampling process at a predetermined frequency for a predetermined time interval, resulting in an ordered set of samples that we refer to as a “sampling vector”.

[0053] In the case we are interested in, i.e. waveform measurements, there are usually two dimensions, namely amplitude and time. As a rule, the representation of a waveform sees amplitude on the vertical axis and time on the horizontal axis, and the expression “resolution” indicates the number of bits used to represent the amplitude of a waveform.

[0054] Sampling a waveform, i.e. the variation in time of a certain parameter, consists of recording a value of this variation every given time interval, the resolution in time being determined by the frequency with which these “samples” are acquired. This frequency is called the “sampling frequency”.

[0055] The sampling step can take place continuously or, more preferably, at regular time intervals. In the second case, sampling is performed by means of the timer of the acquisition device 30 which switches one or more transducers on and off at regular time intervals.

[0056] In the vast majority of practical applications, there is no need for continuous data acquisition. For example, in activities requiring the acquisition of vibration data for predictive maintenance, monitoring with portable devices occurs in most cases at intervals of a few weeks. Thus, sampling the signals of interest every 30 to 60 minutes provides more than adequate knowledge of the phenomena.

[0057] In other embodiments, the monitoring system also includes one or more “trigger” devices. According to these embodiments, the sampling of signals of interest can be carried out on the basis of a specific event. For example, there are cases where it is necessary to recognise the characteristic of a certain sound when it is emitted, with the possibility of discerning different sounds. In this case, sampling must be carried out when a sound of appropriate intensity is picked up by the device.

[0058] According to further embodiments, the data acquisition and processing device, or one or more transducers, must be able to be triggered by a predetermined event but only during predetermined “attention” time windows occurring at regular intervals.

[0059] If the trigger is active, the device will, when the timer activates the time window, wait for the data acquisition and processing device 30 to detect either a given intensity signal or an external trigger signal, and start the sampling sequence of the signals of interest as soon as this event occurs.

[0060] If the trigger is not active, then the device will start sampling immediately upon activation of the time window.

[0061] After the end of the sequence, the device will “switch off”, with the exception of the timer, for a given time interval.

[0062] According to a preferred embodiment of the present invention, the acquisition of the signals of interest is carried out with a sampling process performed in such a way as to make the measurement as repeatable and homogeneous as possible, taking into account that the purpose is to perform a spectrum analysis of a sampled waveform.

[0063] In this regard, and remembering that a Fourier transform must be applied to the analysis of a wave spectrum, two aspects should be emphasised. Firstly, it is in the nature of the Fourier transform that a small shift on the frequency axis of the signal can correspond to an even very large shift on the amplitude axis. Furthermore, it must be remembered that by using a microcontroller and thus a digital device, the result will not be continuous, but discrete and characterised by a frequency resolution of: sampling frequency number of samples For many applications, IOT devices are used to assess how certain signals vary over time, while also trying to identify their causes, through spectrum analysis with complex statistical systems or artificial intelligence. In these cases, waveform sampling has the practical purpose of monitoring the development and evolution of machines, systems or other phenomena over time. Therefore, the assumption is that there is a kind of homogeneity of behaviour in which variations must be identified. A clear example is the sampling of vibrations in a rotary machine in order to prevent major failures or production stoppages.

[0064] It is therefore assumed that the sampled signals must always be somewhat similar if they come from the same predetermined measuring point. Vibrations coming from a predetermined position on a certain machine, sounds coming from the same air or gas flow, and so on.

[0065] For the above reasons, in order to give rise to repeatable measurements and spectrum analysis, the sampling process must firstly involve the acquisition of a fixed number of values, at a given interval for a given time, so that the number of values and the interval are always the same. Since the objective is to assess the variation of signals over time, it is important that the signals themselves are therefore acquired in a homogeneous manner.

[0066] With particular reference to the figures, Fig. 1 a depicts amplitude variations of a signal on a time scale. The signal of interest is somewhat repetitive or, at any rate, has repetitive components of interest. Vibrations on a rotary machine are a good example.

[0067] Figs. 1 b and 1 c depict sampling (vertical lines) carried out in two time intervals of equal duration and subsequent to each other, but occurring at different times with respect to the development of the phenomenon under analysis. It is therefore clear that although the signal does not change over the time interval analysed, the two different sampling intervals, although of equal duration, can produce significantly different results in terms of RMS and spectrum.

[0068] Depending on the preferred embodiment, the sampling of one or more signals of interest may be “phase sampling”.

[0069] “Phase sampling” refers to a process whereby the sampling intervals are chosen in such a way that they are as homogeneous as possible, which allows the sampling intervals to be applied, as illustrated in Fig. 2a and 2b.

[0070] The phase sampling process comprises the following steps:

[0071] (i) establishing a number of samples to be acquired in a sampling vector.

[0072] (ii) (ii) acquiring at a first instant a first sample of a single measurement of the signal of interest and storing the scalar magnitude of said first sample;

[0073] (iii) (iii) identifying the absolute value of said scalar magnitude as the maximum value;

[0074] (iv) (iv) acquiring at a further instant following the previous instant a further sample of a single measurement of the signal of interest and store the scalar magnitude of said further sample;

[0075] (v) (v) comparing the absolute value of the scalar magnitude of said further sample with the maximum value;

[0076] (vi) (vi) if the absolute value of the scalar magnitude of said further sample is less than the maximum value, storing the square of the scalar magnitude of said further sample as the maximum value and return to step (iii);

[0077] (vii) (vii) if the absolute value of the scalar magnitude of said further sample is greater than the maximum value, identifying the scalar magnitude of said further sample as the maximum value, deleting the previously acquired values and returning to step (iii); and

[0078] (viii) (viii) repeating the previous steps until a sampling vector containing the predetermined number of values has been obtained.

[0079] In this way, the sampling obtained always has the maximum measurement acquired as its starting point, allowing the sampling intervals to be applied homogeneously with respect to the shape of the signal.

[0080] Conversely, this first step of the method can consist of a simple sampling operation of a known type, where no repetitive characteristics of the signal of interest to be sampled are expected.

[0081] At the end of the sampling process, we obtain a vector of given size representing the source signal, which will henceforth be referred to simply as the “sampling vector”.

[0082] According to other embodiments, the method of the present invention comprises a step in which a low-pass filter is applied to the sampling vector data.

[0083] Recalling the Nyquist - Shannon theorem and its consequences in application practice, we know that where the sampling frequency used is less than twice the maximum frequency present in the signal, the calculated spectrum may contain “aliasing” phenomena, undesirable for analysis.

[0084] Consequently, if the method is to be applied under the condition that the signal may contain elements of a frequency greater than half the sampling frequency, a low-pass filter must be applied to the sampling vector data.

[0085] The filter can be implemented by known techniques through the introduction of a hardware component at the transducer output of the data acquisition and processing device or through digital transformation of the obtained signal. The waveform acquisition and management method of the present invention comprises a second step in which an RMS of all the data sampled in the previous step, and forming part of the sampling vector, is calculated. RMS is defined as the square root of the mean of the squares of all the data included in the sampling vector. This RMS value is stored in the memory system and is used in the subsequent fractional coding and bit compression coding steps.

[0086] It should be made clear that the calculation of the RMS value can also be performed at a later point in the method, provided that the original values obtained from sampling the waveform are available.

[0087] The microcontrollers used for IOT and HOT devices usually have little memory. Consequently, the practical implementation of methods that can work on widespread and low-cost devices includes the need to optimise the use of available memory as much as possible. The vector of sampled values is a rather large memory element, which includes, precisely, hundreds or thousands of values. Calculating the RMS at a later time would require preserving the original values and the original vector, creating one of equal size for subsequent calculations.

[0088] By calculating the RMS at this point in the method and storing its value in a specific variable, subsequent calculations on the sampling vector values can be performed in the same memory space, without storing the original values.

[0089] Furthermore, it should be considered that even if controllers with more memory than the current ones were available in the future, it would still be advantageous to use this memory to obtain more precise measurements and better resolutions. According to some embodiments, ancillary values obtainable from the sampled data may be calculated at this stage. In the case of vibration, for example, speed and displacement. With regard to the time at which these calculations are performed in the method, the same applies as for the calculation of the RMS value.

[0090] The waveform acquisition and management method of the present invention comprises a third step in which the sampling vector obtained in the first step is processed using the discrete Fourier transform (FFT) to obtain a spectrum, the spectrum itself being represented by a vector, referred to as the “FFT vector”.

[0091] The choice of which data to keep from the FFT vector for the next step depends on the actual application of the system. It must be remembered that the Fourier transform is a symmetrical function. Furthermore, the Fourier transform of a symmetrical function is itself symmetrical.

[0092] For this reason, the first half of the data of an FFT of a signal that we assume to be symmetrical contains exactly the same information as if we included the second half, which can lead to considerable data savings.

[0093] Thus, the actual implementation of the algorithm must take into account whether the data we are interested in can be represented with a symmetrical function or not.

[0094] If the data we are interested in are represented by a symmetrical function, due to the nature of the phenomenon being observed, then the FFT vector will only contain the first half of the data obtained from the FFT.

[0095] If we can expect that the acquired data cannot be represented by a symmetrical function, then the FFT vector will include all data obtained from the FFT calculation.

[0096] According to a possible embodiment of the present invention, FFT vector data is stored in the memory system by overwriting previously stored sampling vector data, optimising the space occupied in the memory system. The need to save memory is in fact a crucial element, as highlighted above.

[0097] However, the FFT vector includes a large number of values and thus data, as sampling is normally carried out by sampling several thousand values. However, the size of messages that can be sent with Lo.Ra.-like vectors is only a few dozen characters.

[0098] The amount of bits required to store a certain value depends on the range that this value can take. Thus, 16 bits, or two bytes, are required to represent the numbers 0 to 65535. To represent a number from 0 to 255, 8 are sufficient. Thus, it is evident that the range of values that each variable can take in a calculation has a decisive influence on the amount of memory required to accommodate and transfer its results. The messages used with communication technologies similar to Lo.Ra., are usually “strings”, i.e. sequences of characters, where each character can occupy 1 byte, i.e. 8 bits.

[0099] Therefore, once the number of values that the communication vector is actually capable of transferring is known, it is necessary to reduce the frequency resolution of the spectrum appropriately in order to obtain a “reduced spectrum”.

[0100] The waveform acquisition and management method of the present invention thus comprises a fourth step of reducing the frequency resolution of the FFT vector data.

[0101] There are various ways to reduce frequency resolution. The simplest is to average the amplitudes for homogeneous groups. Therefore, in the case of a 1 Hz resolution and wanting to obtain a 10Hz resolution, the spectrum is divided into groups of 10 values each and each interval is averaged to obtain the final result. This already well-known and widespread method has the major flaw of breaking down isolated peaks to a very great extent, and thus not adequately “preserving” the geometric shape of the spectrum, and in some cases could lead to errors in the identification of dominant frequencies. This flaw is extremely important as it is definitely not suitable for the main purpose of the present invention, which is predictive analysis.

[0102] Otherwise, one of the purposes of the present invention is to adequately preserve the geometric shape of the spectrum while reducing the size of the message used to transfer the result of a sampling operation to a minimum.

[0103] According to a preferred embodiment of the present invention, frequency reduction comprises calculating the value of each interval as the sum of the values of the coefficients included in that interval that exceed the average of the values of the spectrum.

[0104] The waveform acquisition and management method of the present invention thus comprises a fifth step of “fractional recoding”.

[0105] In view of the fact that the RMS value of the acquired signal has been calculated, i.e. the RMS, it may be convenient to transfer the respective values of the previously reduced spectrum as numbers between 0 and 1000, or other denominator, calculated in proportion to the original values by performing this approximation with an algorithm that allows us to obtain a vector of values, the sum of which will be preserved equal to one thousand.

[0106] Fractional recoding aims to reduce the amount of data required to save or transfer useful information, according to a known approximation term.

[0107] Fractional recoding according to the present invention comprises a first step in which a predetermined divisor value, referred to herein as the “denominator”, is established. The value is set depending on the desired output resolution and the amount of data available for storage or individual transfer, so as to keep the data to be transferred as small as possible while ensuring that the approximation and thus the accuracy of the data is acceptable. In the preferred embodiment, the divisor is a multiple of 10.

[0108] The reduced spectrum, to be transferred or saved, is a vector composed of the following elements:

[0109] - previously calculated RMS value, and

[0110] - a plurality of vector elements each calculated as follows:

[0111] The following must be taken into account when choosing the denominator.

[0112] As a rule, messages for data transfer on long-range vectors such as Lo.Ra. are string-based.

[0113] This type of representation, in terms of memory occupation, and thus the actual volume of data to be transferred, proves to be extremely disadvantageous. Using, for example, decimal numbering, considering that each ASCII character occupies 8 bits, 32 bits would be used to represent the number 1000 as a string. The same number, for simple binary encoding, requires 10 bits.

[0114] The choice of denominator will determine the number of bits required for each element of the final data to be transferred and will also determine the level of approximation with which the data will be transferred.

[0115] Furthermore, theoretically, the sum of the results obtained from the respective elements of the reduced spectrum, according to the above formula, should equal the chosen denominator. However, by simply performing the numerical calculation as shown above, the individual approximations do not lead to this result and are not easy to control. Thus, in a preferred version of implementing the method, the calculation of value elements for fractional recoding includes an algorithm for optimal handling of approximations.

[0116] According to this algorithm, each value of the fractional encoding of the reduced spectrum is calculated as follows, and on the basis of the following definitions:

[0117] - SumSR = the sum of the values (real numbers) of the reduced spectrum to be encoded, calculated in the frequency resolution reduction step;

[0118] - tempReal = a real number to be saved during the algorithm for subsequent steps;

[0119] - tempinteger = an integer to be saved during the algorithm for subsequent steps;

[0120] - VectorSR[i] = ordered set of reduced spectrum values to be encoded;

[0121] - VectorCF[i] = ordered set of values obtained as a result of fractional encoding.

[0122] For each i-th element of VectorSR[i] if i = 0

[0123] - VectorCF[i] = integer part of ((denominator*VectorSR[i] / sumSR)+0.5)

[0124] - tempReal = denominator*VectorSR[i] / SumSR

[0125] - tempinteger = VectorCF[i]; if i > 0

[0126] - tempReal = tempReal + Denominator*VectorSR[i] / SumSR

[0127] - VectorCF[i] = Integer part of (tempReal+0.5) - tempinteger;

[0128] - tempinteger = tempinteger + VectorCF[i], The method of the present invention comprises a sixth step of bit encoding of the vectorCF data, obtained as a result of the fractional encoding applied in the previous step.

[0129] Once the spectrum has been obtained reduced into fractions with the denominator previously established (VectorCF), according to the step indicated above, it is necessary to proceed with the encoding of the data, transforming it appropriately into a message that occupies as few bytes as possible.

[0130] As a result of this step, the RMS value and ancillary values may be encoded as “floats”, i.e. they may each occupy a group of 32 bits at the beginning, end or at an a priori known position in the message that will be obtained as a result of this step.

[0131] The message will then be composed first in bits, using 32-bit spaces for the RMS and each other accessory value. Subsequently, spaces composed of the minimum number of bits required to represent the denominator will be used in order to contain each number of the vectorCF.

[0132] This seventh step consists of the following steps:

[0133] - identifying a maximum number N of bits to represent the denominator chosen for the execution of the previous step;

[0134] - - having said I as the number of elements contained in the VectorCF, create a vector of individual bits VectorBit containing I x N bits;

[0135] - - for each value of the VectorCF, expressed by integers, copying the N useful bits into the VectorBit; depending on the system in use, the useful bits may be the first N (big endian) or the last N (little endian) bits;

[0136] - - adding to the beginning or end of the VectorBit all the bits of the real number RMS, which in this way is transferred without approximations; and - - dividing the bits contained in the VectorBit into groups of 8 bits, so that each group becomes the character of a string, which can then be transferred or saved, with very considerable savings in data.

[0137] The resulting message can be sent using the appropriate Lo.Ra.-like vector.

[0138] If there are several transducers in the data acquisition and processing device, the process described above is repeated for each transducer.

[0139] The method of the present invention comprises a seventh step in which data thus processed by said one or more devices for automatic data acquisition and / or storage and / or analysis are transmitted to a server or computer 10 by means of a radio transmission system employing a long-range vector.

[0140] In other embodiments, the method includes a further step of decoding the encoded product spectrum data, which makes them effectively fungible and usable on standard information systems

[0141] The string obtained in the previous step can be saved on the data acquisition and processing device, or, as mentioned above, transmitted to the server via the data transmission device and the antennas / gateways. In both cases, the string must be decoded using a decoding algorithm.

[0142] Said algorithm comprises a first step of:

[0143] - determining how much ancillary data beyond the RMS has been inserted into the message and whether it has been placed, assuming the known order, at the beginning or end of the string and how many bits each data occupies;

[0144] - - determining whether the encoding is big endian or little endian;

[0145] - - determining the denominator and the respective number N;

[0146] - - determining the number of elements I of VectorCF; - determining the number of bits and the type for each numeric data element of vectorCF.

[0147] The algorithm also includes the steps of:

[0148] - creating a bit vector VectorBit that is filled with all the bits in the string;

[0149] - - creating a vector VectorCFdec of numeric data of suitable type with I elements;

[0150] - - writing the ith data element of VectorCFdec by taking a number of bits equal to N o - from position i*N of the bit vector, if RMS is at the end, o from position i*N + (number of bits occupied by the RMS), if the RMS is at the beginning.

[0151] - adding enough zeros to these bits to reach the number of bits of the original data type at the beginning, if the encoding is little endian, or at the end, if the encoding is big endian; and

[0152] - - writing the data obtained in the original format to the ith position of VectorCFdec.

[0153] Note the fact that where the transmitting and receiving machines have different endianesses, this must be taken into account by appropriately reversing the bit order.

[0154] The use of the system and method described here, with the device shown, makes monitoring installations with thousands of connected devices feasible and sustainable.

[0155] According to other embodiments of the present invention, the system for acquiring and managing waveforms on IOT and HOT devices and systems for spectrum analysis includes transducers on board acquisition devices that lack a predetermined “O-position” value on value 0, such as MEMS accelerometers.

[0156] When using a MEMS-type accelerometer, the axis origin at which the acquired signal amplitude variations are evaluated varies depending on the position itself with respect to the ground plane. This is because the sensor is able to measure the acceleration of gravity. Thus, when the sensor axis is perpendicular to the ground, the sensor will report an acceleration value of 1 G, which will be the axis origin for the detection of any vibrations perpendicular to the ground. For example, the 0 of an ADXL345 accelerometer, i.e. the value that the component measures when stationary with the z-axis perpendicular to the ground, is worth about 256 on that axis.

[0157] Requiring the position of MEMS in the field to be perfectly homogeneous is not possible from a practical point of view. In some cases, the actual position of the transducer may not be accurate or known. In other cases, the sensor signal may be “shifted” for reasons of expediency related to the design of the electronic circuit.

[0158] Consequently, it is necessary to “neutralise” the possible dislocation of the origin of the axes.

[0159] In these cases, after the wave sampling step, the average value of all values is subtracted from each acquired value.

[0160] We can therefore say that the subtraction of the average of the values acquired from each value applies where the natural origin of the axes of measurement has, for at least one of the two axes, a value other than 0. Note the fact that this step changes the FFT results very little, but modifies the calculation of the RMS, on which the reconstruction of the output data then depends.

[0161] The use of the system and method described here, with the device shown, makes monitoring installations with thousands of connected devices feasible and sustainable.

Claims

Claims1. A method of waveform acquisition and management on IOT and HOT devices and systems aimed at a spectrum analysis includes the following steps:- sampling according to regular time intervals one or more signals from a physical emitter (40) by means of one or more data acquisition and processing devices (30) and obtaining a sampling vector;- processing the sampling vector data through said data acquisition and processing unit (30) by calculating the RMS of the sampled data;- processing the sampling vector data through said control and data processing unit (30) by performing a Fourier Transform calculation of the sampling vector and obtain an FFT vector representing a spectrum of the one or more sampled signals;- performing a frequency resolution reduction of the FFT vector data through said control and data processing unit (30) and obtain a reduced spectrum of the one or more sampled signals;- performing fractional recoding of the data obtained from the previous step through said data processing and control unit (30) to reduce the number of data needed to transfer the useful information;- performing bit compression coding of the data from the spectrum reduced in the previous step through said data control and processing unit (30); and- transmitting the data thus processed to a data processor (10) by means of a radio transmission system (20) employing a long-range, low-power transmission mode.

2. Method according to claim 1 , characterized in that the sampling step comprises the steps of:(i) (i) establishing a number of samples to be acquired in a sampling vector;(ii) (ii) acquiring at a first instant a first sample of a single measurement of the signal of interest and storing the scalar magnitude of said first sample;(iii) (iii) identifying the absolute value of said scalar magnitude as the maximum value;(iv) (iv) acquiring at a further instant following the previous instant a further sample of a single measurement of the signal of interest and store the scalar magnitude of said further sample;(v) (v) comparing the absolute value of the scalar magnitude of said further sample with the maximum value;(vi) (vi) if the absolute value of the scalar magnitude of said further sample is less than the maximum value, storing the square of the scalar magnitude of said further sample as the maximum value and return to step (iii);(vii) (vii) if the absolute value of the scalar magnitude of said further sample is greater than the maximum value, identifying the scalar magnitude of said further sample as the maximum value, deleting the previously acquired values and returning to step (iii); and(viii) (viii) repeating the previous steps until a sampling vector containing the predetermined number of values has been obtained.

3. Method according to any of the preceding claims, characterized in that the fractional recoding step comprises the steps of:- determining an ordered set of values (VectorCF[i]), a first variable (tempReal), and a second variable (tempinteger), where for each i-th element of VectorSR[i] if i = 0- - VectorCF[i] = integer part of ((Denominator*VectorSR[i] / sumSR)+0.5),- - tempReal = denominator*VectorSR[i] / SumSR, and- - tempinterval = VectorCF[i], if i > 0:- - VectorCF[i] = Integer part of (tempReal+0.5) - tempinteger;- - tempReal = tempReal + Denominator*VectorSR[i] / SumSR- - tempinteger = tempinteger +VectorCF[i] where:- denominator = predetermined divisor value set according to the desired output resolution and the amount of data available for transmission;- SumSR = sum of the reduced spectrum values to be encoded, calculated in the frequency resolution reduction step; and- VectorSR[i] = ordered set of the reduced spectrum values to be encoded.

4. A method according to any of the preceding claims, characterized in that between the sampling step and the RMS calculation step includes a step in which the average value of all acquired values is subtracted from each acquired value.

5. Method according to any of the preceding claims, characterized in that the bit encoding step of the reduced spectrum data includes the steps of:- - identifying a maximum number N of bits to represent the denominator chosen for the execution of the previous step;- - having said I as the number of elements contained in the VectorCF, create a vector of individual bits VectorBit containing I x N bits;- - for each value of the VectorCF, expressed by integers, copying the N useful bits into the VectorBit; depending on the system in use, the useful bits may be the first N (big endian) or the last N (little endian) bits;- - adding to the beginning or end of the VectorBit all the bits of the real number RMS, which in this way is transferred without approximations; and- - dividing the bits contained in the VectorBit into groups of 8 bits, so that each group becomes the character of a string, which can then be transferred or saved, with very considerable savings in data.

6. A method according to any of the preceding claims, characterized in that it further includes a step of decoding the encoded produced spectrum data.

7. Method according to claim 6, characterized in that the decoding step comprises the steps of:- - determining how much ancillary data beyond the RMS has been inserted into the message and whether it has been placed, assuming the known order, at the beginning or end of the string and how many bits each data occupies;- - determining whether the encoding is big endian or little endian;- - determining the denominator and the respective number N;- - determining the number of elements I of VectorCF;- - determining the number of bits and type for each numeric data element of VectorCF;- - creating a bit vector VectorBit that is filled with all the bits in the string;- - creating a vector VectorCFdec of numeric data of suitable type with I elements;- - writing the ith data element of VectorCFdec by taking a number of bits equal to N- - from position i*N of the bit vector, if RMS is at the end,- - from position i*N + (number of bits occupied by RMS), if RMS is at the beginning;- - adding enough zeros to these bits to reach the number of bits of the original data type at the beginning, if the encoding is little endian, or at the end, if the encoding is big endian; and- - writing the data obtained in the original format to the ith position of VectorCFdec.

8. A method according to any of the preceding claims, characterized in that after the sampling phase includes a phase in which to apply a low-pass filter to the sampling vector if the sampling frequency is less than twice the maximum frequency of the signal.

Images