{"id":2138,"date":"2023-04-07T21:01:25","date_gmt":"2023-04-07T21:01:25","guid":{"rendered":"https:\/\/vibromera.eu\/?p=2138"},"modified":"2026-07-02T19:45:53","modified_gmt":"2026-07-02T19:45:53","slug":"application-of-the-fourier-transform-to-the-analysis-of-vibration-signals","status":"publish","type":"post","link":"https:\/\/vibromera.eu\/sw\/example\/application-of-the-fourier-transform-to-the-analysis-of-vibration-signals\/","title":{"rendered":"Application of the Fourier transform to the analysis of vibration signals."},"content":{"rendered":"

Matumizi ya Mabadiliko ya Fourier katika Uchambuzi wa Ishara za Mtetemo<\/h1>\n

Andrei Shelkovenko. One of the developers and founder of Vibromera.
\nThe translation of the article may contain inaccuracies.<\/p>\n

Fourier transform and signal spectrum<\/h2>\n

Katika hali nyingi, kazi ya kupata (kuhesabu) spectrum<\/a> ya ishara ni kama ifuatavyo. Kuna ADC, ambayo kwa sampuli ya frequency<\/a> Fd inabadilisha ishara inayoendelea, inayoingia kwake wakati wa muda T, kuwa sampuli za dijiti – N vipande. Kisha safu hii ya sampuli inalishwa kwenye programu fulani (kwa mfano FourierScope<\/a><\/span>) which outputs N\/2 some numeric values.<\/p>\n

To check if the program works correctly, we form an array of samples as a sum of two sin(10*2*pi*x)+0.5*sin(5*2*pi*x) and feed it into the program. The program drew the following:
\n

\"Fourier

Fig.1 The graph of the time function of the signal<\/span><\/i><\/em><\/p><\/div><\/p>\n

 <\/p>\n

\"Fig.2

Fig.2 The graph of the signal spectrum<\/p><\/div>\n

 <\/p>\n

There are two harmonics<\/a> kwenye grafu ya wigo – 5 Hz na amplitude ya 0.5 V na 10 Hz na amplitude ya 1 V, kila kitu ni kama katika fomula ya ishara asili. Kila kitu kiko sawa, programu inafanya kazi ipasavyo.<\/p>\n

This means that if we feed a real signal from a mixture of two sinusoids to the ADC input, we will get a similar spectrum consisting of two harmonics.<\/p>\n

So, our real <\/span><\/b><\/strong>measured signal of 5 sec. duration<\/span><\/b><\/strong>, digitized by ADC, i.e. represented by discrete <\/span><\/b><\/strong>samples, has a discrete non-periodic <\/span><\/b><\/strong>spectrum.
\nFrom a mathematical point of view \u2013 how many errors in this phrase? <\/span><\/i><\/em><\/p>\n

Now let\u2019s try to measure the same signal for 0.5 sec.<\/p>\n

\"Fig.3

Fig.3 The graph of the function sin(10*2*pi*x)+0.5*sin(5*2*pi*x) for a measurement period of 0.5 sec<\/p><\/div>\n

 <\/p>\n

\"Fig.4

Fig.4 Spectrum of the function<\/p><\/div>\n

 <\/p>\n

Something is wrong here! The harmonic at 10 Hz is drawn normally, and instead of the harmonic at 5 Hz there are some unclear harmonics.<\/p>\n

On the Internet they say that it is necessary to add zeros to the end of the sample and the spectrum will be drawn normally.<\/p>\n

\"Fig.5

Fig.5 We have added zeros to the sample up to 5 sec<\/p><\/div>\n

 <\/p>\n

\"Fig.6.

Fig.6. Spectrum obtained.<\/p><\/div>\n

 <\/p>\n

That\u2019s not it at all. I will have to deal with the theory. Let\u2019s go to wikipedia<\/b><\/strong><\/a><\/span>\u00a0\u2013 the source of knowledge.<\/p>\n

Continuous function and its Fourier series representation<\/h2>\n

Mathematically, our signal with duration of T seconds is some function f(x) given on the interval {0, T} (X in this case is time). Such a function can always be represented as a sum of harmonic functions (sine or cosine) of the form:<\/p>\n

\"Continuous

\u00a0(1), where:<\/p>\n

<\/p><\/div>\n

k is the number of the trigonometric function ( the number of the harmonic component, the number of the harmonic)
\nT \u2013 segment where the function is defined (the duration of the signal)
\nAk- amplitude of the k-th harmonic component,
\n\u03b8k- the initial phase of the kth harmonic component
\nWhat does it mean to \u201crepresent the function as the sum of the series\u201d? It means that by adding the values of the harmonic components of the Fourier series at each point, we get the value of our function at that point.
\n(More strictly, the mean square deviation of the series from the function f(x) will tend to zero, but despite the mean square convergence, the Fourier series of a function need not, generally speaking, converge to it point by point. )
\nThis series can also be written in the form:<\/p>\n

\"(2),\"

(2),<\/p><\/div>\n

 <\/p>\n

 <\/p>\n

 <\/p>\n

where \"Mlingano , the kth complex amplitude.<\/p>\n

 <\/p>\n

or<\/p>\n

\"

(3)<\/p><\/div>\n

 <\/p>\n

 <\/p>\n

 <\/p>\n

The relationship between the coefficients (1) and (3) is expressed by the following formulas:<\/p>\n

\"Fomula<\/p>\n

 <\/p>\n

 <\/p>\n

\"Fomula<\/p>\n

 <\/p>\n

 <\/p>\n

Note that all these three representations of Fourier series are perfectly equivalent. Sometimes when working with Fourier series, it is more convenient to use exponents of imaginary argument instead of sines and cosines, i.e. to use Fourier transform in complex form. But it is convenient for us to use formula (1), where the Fourier series is represented as a sum of cosines with corresponding amplitudes and phases. Strictly speaking, the Fourier transform of a real signal does produce complex coefficients (form (3)): each coefficient carries both the amplitude and the phase of its harmonic. For a real signal these complex coefficients possess conjugate (Hermitian) symmetry \u2014 the negative-frequency half simply mirrors the positive-frequency half and carries no extra information. That is why, from the complex coefficients, we can always pass to the real non-negative amplitudes Ak and phases \u03b8k of formula (1) \u2014 and it is exactly this amplitude spectrum that analysis programs plot.<\/p>\n

 <\/p>\n

Bottom line:<\/span><\/b><\/u><\/strong>
\n<\/span><\/b><\/strong>The mathematical basis for spectral analysis of signals is the Fourier transform.<\/span><\/b><\/strong>
\n<\/span><\/b><\/strong>
\n<\/span><\/b><\/strong>The Fourier transform allows to represent a continuous function f(x) (signal) defined on the interval {0, T} as a sum of infinite number (infinite series) of trigonometric functions (sine and\/ or cosine) with definite amplitudes and phases also considered on the interval {0, T}. Such series is called a Fourier series.<\/span><\/b><\/strong><\/p>\n

Note some more points, the understanding of which is required for correct application of the Fourier transform to signal analysis. If we consider the Fourier series (sum of sinusoids) on the entire X-axis we will see that outside the interval {0, T} the Fourier series function will periodically repeat our function.<\/p>\n

For example, in the graph in Fig. 7, the original function is defined on the interval {-T\\2, +T\\2}, and the Fourier series represents a periodic function defined on the entire x-axis.<\/p>\n

This is because the sinusoids themselves are periodic functions, so their sum will also be a periodic function.<\/p>\n

\"Figure

Figure 7 Representation of a non-periodic source function by a Fourier series<\/p><\/div>\n

Thus:<\/p>\n

Our original function is a continuous, non-periodic function defined on some segment of length T.
\nThe spectrum of this function is discrete, i.e. it is represented as an infinite series of harmonic components \u2013 a Fourier series.
\nIn fact, Fourier series defines some periodic function, which coincides with our function on the interval {0, T}, but for us this periodicity is not essential.<\/p>\n

Next.<\/p>\n

The periods of harmonic components are multiples of the interval {0, T}, on which the initial function f(x) is defined. In other words, the periods of harmonics are multiples of the duration of the signal measurement. For example, the period of the first harmonic in a Fourier series is equal to the interval T in which the function f(x) is defined. The period of the second harmonic in a Fourier series is equal to the interval T\/2. And so on (see Figure 8).<\/p>\n

\"Fig.

Fig. 8 Periods (frequencies) of harmonic components of the Fourier series (here T=2\u03c0)<\/p><\/div>\n

Accordingly, the frequencies of harmonic components are multiples of 1\/T. That is, frequencies of harmonic components Fk are Fk= k\\T, where k runs values from 0 to \u221e, for example, k=0 F0=0; k=1 F1=1\\T; k=2 F2=2\\T;k=3 F3=3\\T;\u2026. Fk= k\\T (at zero frequency, a constant component).<\/p>\n

Let our initial function, is a signal recorded during T=1 sec. Then the period of the first harmonic will be equal to the duration of our signal T1=T=1 sec and the frequency of the harmonic is equal to 1 Hz. The period of the second harmonic will be equal to the duration of our signal divided by 2 (T2=T\/2=0.5 sec.) and the frequency is equal to 2 Hz. For the third harmonic, T3=T\/3 sec and the frequency is 3 Hz. And so on.<\/p>\n

The step between harmonics in this case is 1 Hz.<\/p>\n

Thus, a signal with a duration of 1 sec can be decomposed into harmonic components (to obtain a spectrum) with a frequency resolution of 1 Hz.
\nIli kuongeza azimio mara 2 hadi 0.5 Hz, ni lazima kuongeza muda wa kipimo mara 2 hadi sekunde 2. Ishara ya sekunde 10 inaweza kugawanywa katika vipengele vya harmonic (wigo) na azimio la masafa ya 0.1 Hz. Hakuna njia nyingine za kuongeza azimio la masafa. Unaweza kuchunguza uhusiano huu na FFT Resolution Calculator<\/a>.<\/p>\n

There is a way to artificially increase the signal duration by adding zeros to the array of samples. But it does not increase the real frequency resolution.<\/p>\n

Ishara tofauti na mabadiliko tofauti ya Fourier<\/h2>\n

With the development of digital technology the ways of measurement data (signals) storage have changed. Whereas previously a signal could be recorded on a tape recorder and stored on a tape in analog form, now signals are digitized and stored in files in computer memory as a set of numbers (counts).<\/p>\n

The usual scheme of signal measurement and digitization looks as follows.<\/p>\n

Measuring transducer \u2014- Signal normalizer \u2014- ADC \u2014\u2013 Computer
\n(Fig.9 Schematic of the measuring channel)<\/p>\n

<\/span><\/i><\/em><\/p>\n

The signal from the measuring transducer goes to the ADC for a period of time T. The signal readings (sampling) received during time T are transmitted to the computer and saved in the memory.<\/p>\n

\"Fig.10

Fig.10 Digitized signal \u2013 N samples received for time T<\/p><\/div>\n

What are the requirements to signal digitization parameters? A device which converts the input analog signal into a discrete code (digital signal) is called an analog-to-digital converter (ADC) (\u00a9 Wiki).<\/p>\n

One of the basic parameters of ADC is the maximum sampling rate \u2013 the frequency of sampling of a signal which is continuous in time. Sample rate is measured in hertz. ((\u00a9 Wiki))<\/p>\n

According to Kotelnikov’s theorem, if a continuous signal has a spectrum limited by the frequency Fmax, it can be fully and uniquely reconstructed from its discrete samples taken at time intervals\u00a0\u0394t \u2264 1\/(2*Fmax), ie with a sampling frequency Fd \u2265 2*Fmax, where Fd – sampling frequency; Fmax – the maximum frequency of the signal spectrum. In other words, the frequency of signal digitization (sampling frequency of ADC) must be at least twice the maximum frequency of the signal we want to measure.<\/p>\n

And what will happen if we take samples with lower frequency than required by Kotelnikov\u2019s theorem?<\/p>\n

Katika hali hii kuna “aliasing<\/a>In this case there is an \u201caliasing\u201d effect (aka stroboscopic effect, moir\u00e9 effect), in which a high frequency signal after digitization turns into a low frequency signal, which in fact does not exist. In Fig. 11 the red sine wave of high frequency is the real signal. The blue sine wave of lower frequency is a fictitious signal, arising due to the fact that during the sampling time has time to pass more than half a period of the high-frequency signal.<\/p>\n

\"Fig.

Fig. 11. Appearance of a spurious low frequency signal at an insufficiently high sampling rate<\/p><\/div>\n

 <\/p>\n

Ili kuepuka athari ya aliasing, kichujio maalum cha kuzuia-aliasing (low-pass filter<\/a>) kinawekwa mbele ya ADC. Kinapitisha masafa chini ya nusu ya masafa ya sampuli ya ADC na kukata masafa ya juu zaidi.<\/p>\n

Ili kuhesabu wigo wa ishara kutoka kwa sampuli zake tofauti, mabadiliko tofauti ya Fourier (DFT)<\/a> hutumiwa. Kumbuka tena kwamba wigo wa ishara tofauti ni “kwa ufafanuzi” umezuiliwa kwa masafa Fmax ndogo kuliko nusu ya masafa ya sampuli Fd. Kwa hivyo, wigo wa ishara tofauti unaweza kuwakilishwa kwa jumla ya a finite <\/u>number of harmonics, in contrast to the infinite sum for the Fourier series of a continuous signal, whose spectrum can be unlimited. According to Kotelnikov\u2019s theorem, the maximum frequency of a harmonic must be such that it accounts for at least two samples, so the number of harmonics is equal to half the number of samples of a discrete signal. That is, if there are N samples in the sample, the number of harmonics in the spectrum will be N\/2.<\/p>\n

Consider now the discrete Fourier transform (DFT).<\/p>\n

\"Equation<\/p>\n

Comparing it with the Fourier series<\/p>\n

 <\/p>\n

\"Fomula<\/p>\n

As we can see, they coincide, except for the fact that time in the FFT is discrete and the number of harmonics is limited to N\/2, which is half the number of samples.<\/p>\n

DFT formulas are written in dimensionless integer variables k, s, where k is the number of signal samples, s is the number of spectral components.
\nThe value s shows the number of full harmonic oscillations per period T (signal measurement duration). The discrete Fourier transform is used to find the amplitudes and phases of harmonics numerically, i.e. \u201con the computer\u201d.<\/p>\n

As it was already said above, when decomposing a non-periodic function (our signal) into Fourier series, the resulting Fourier series actually corresponds to a periodic function with period T (Fig.12).<\/p>\n

 <\/p>\n

\"Fig.12.

Fig.12. Periodic function f(x) with period T0, with period T>T0<\/p><\/div>\n

 <\/p>\n

Kama inavyoonekana katika Mchoro 12, kazi f(x) ni ya mzunguko na kipindi T0. Hata hivyo, kwa sababu urefu wa sampuli ya kipimo T haufanani na kipindi cha kazi T0, kazi inayopatikana kama mfululizo wa Fourier ina mkatiko katika hatua T. Kwa sababu hiyo, wigo wa kazi hii utakuwa na idadi kubwa ya harmonic za masafa ya juu. Jambo hili linajulikana kama spectral leakage<\/a>, na katika mazoea hupunguzwa kwa windowing<\/a> ishara kabla ya mabadiliko. Kama muda wa sampuli ya kipimo T ungelingana na kipindi cha kazi T0, basi wigo uliopatikana baada ya mabadiliko ya Fourier ungehusisha harmonic ya kwanza tu (sinusoid yenye kipindi kinachofanana na muda wa sampuli), kwa sababu kazi f(x) ni sinusoid.<\/p>\n

In other words, the DFT program \u201cdoes not know\u201d that our signal is a \u201cslice of a sine wave\u201d, but tries to represent as a series a periodic function which has a discontinuity due to the discontinuity of separate pieces of the sine wave.<\/p>\n

As a result, harmonics appear in the spectrum, which should, in total, represent the shape of the function, including this discontinuity.<\/p>\n

Thus, to get a \u201ccorrect\u201d spectrum of a signal which is a sum of several sinusoids with different periods, it is necessary that an integer number of periods of <\/u>each sinusoid should be present on the measuring period of the signal. In practice, this condition can be met with a sufficiently long duration of signal measurement.<\/p>\n

 <\/p>\n

\"Fig.13

Fig.13 Example of the kinematic error signal function and spectrum of a gearbox<\/p><\/div>\n

 <\/p>\n

At shorter duration the picture will look \u201cworse\u201d:<\/p>\n

 <\/p>\n

\"Fig.14

Fig.14 Example of rotor vibration function and spectrum<\/p><\/div>\n

 <\/p>\n

 <\/p>\n

 <\/p>\n

In practice, it can be difficult to understand where the \u201creal components\u201d and where the \u201cartifacts\u201d caused by inconsistency of component periods and signal sampling durations or \u201cjumps and breaks\u201d in the waveform. Of course, the words \u201creal components\u201d and \u201cartifacts\u201d are put in quotes for a reason. The presence of many harmonics on the spectrum graph does not mean that our signal actually consists of them. It is like thinking that number 7 \u201cconsists\u201d of numbers 3 and 4. The number 7 can be thought of as the sum of 3 and 4 \u2013 that is correct.<\/p>\n

So also our signal\u2026 or rather not even \u201cour signal\u201d, but a periodic function composed by repeating our signal (sample) can be represented as a sum of harmonics (sine waves) with certain amplitudes and phases. But in many cases important for practice (see figures above) it is indeed possible to relate the harmonics obtained in the spectrum also to real processes having cyclic character and contributing significantly to the form of the signal.<\/p>\n

Some results<\/h2>\n

1. A real measured signal with T sec. duration digitized by ADC, i.e. represented by a set of discrete samples (N pieces), has a discrete non-periodic spectrum represented by a set of harmonics (N\/2 pieces).<\/p>\n

2. The signal is represented by a set of real values. Its DFT spectrum is a set of complex coefficients with conjugate symmetry; from them the amplitude spectrum is obtained \u2014 a set of real non-negative amplitudes (and phases) at positive frequencies, and it is this one-sided amplitude spectrum that is plotted in practice. The two-sided complex form with negative frequencies and the one-sided amplitude\/phase form are equivalent representations of the same spectrum \u2014 for signal analysis it is usually more convenient to work with the one-sided amplitude spectrum.<\/p>\n

3. The signal measured at time T is determined only at time T. What happened before we started measuring the signal and what will happen after that is unknown to science. And in our case it is not interesting. FFT of the time-limited signal gives its \u201creal\u201d spectrum, in the sense that under certain conditions it allows to calculate the amplitude and frequency of its components.<\/p>\n

 <\/p>","protected":false},"excerpt":{"rendered":"

Application of the Fourier Transform to the Analysis of Vibration Signals Andrei Shelkovenko. One of the developers and founder of Vibromera. The translation of the article may contain inaccuracies. Fourier transform and signal spectrum In many cases the task of obtaining (calculating) the spectrum of a signal is as follows. […]<\/p>\n","protected":false},"author":2,"featured_media":2161,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[4],"tags":[],"class_list":["post-2138","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-example"],"_links":{"self":[{"href":"https:\/\/vibromera.eu\/sw\/wp-json\/wp\/v2\/posts\/2138","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/vibromera.eu\/sw\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/vibromera.eu\/sw\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/vibromera.eu\/sw\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/vibromera.eu\/sw\/wp-json\/wp\/v2\/comments?post=2138"}],"version-history":[{"count":2,"href":"https:\/\/vibromera.eu\/sw\/wp-json\/wp\/v2\/posts\/2138\/revisions"}],"predecessor-version":[{"id":102137,"href":"https:\/\/vibromera.eu\/sw\/wp-json\/wp\/v2\/posts\/2138\/revisions\/102137"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/vibromera.eu\/sw\/wp-json\/wp\/v2\/media\/2161"}],"wp:attachment":[{"href":"https:\/\/vibromera.eu\/sw\/wp-json\/wp\/v2\/media?parent=2138"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/vibromera.eu\/sw\/wp-json\/wp\/v2\/categories?post=2138"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/vibromera.eu\/sw\/wp-json\/wp\/v2\/tags?post=2138"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}