Order Filter

The Order Filter analysis object calculates order tracking for speed-dependent vibrations by bandpass filtering individual orders. Vibrations measured on rotating machines show a spectrum in which maxima occur at frequencies corresponding to a multiple of the machine's speed. There are two different causes for the appearance of these maxima. On the one hand, you can view the machine as a non-linear transmission system that is activated by a harmonic vibration corresponding to the speed. The non-linearity generates harmonic components of this fundamental vibration that lead to the corresponding maxima. On the other hand, this type of machine may contain components whose speed is not equal to the base speed, but always corresponds to a fixed multiple of this speed. Different axles have different speeds in a transmission, for instance. But also the teeth of a gearwheel or the balls in a ball bearing create vibrations that are permanently related to the speed. If you recognize this relationship described as the order from the fundamental frequency of a component to the fundamental frequency of the machine, then individual maxima in a spectrum can be assigned to one or a few components of the machine. This can be used, for example, to isolate the cause of resonances.

In the method used here, a sampling rate conversion of the time signal is first performed (revolution synchronous resampling) so that the signal is no longer present in temporally equidistant steps, but in equidistant rotation angle steps (i.e. equidistant rotation intervals). This provides an effective method for performing order tracking, since ordinary band pass filtering of the signal converted to the revolution domain (i.e. using a suitable IIR band pass filter around the respective order) then provides the individual order curves directly (for details, see RevolutionSyncSampling:

The Order Filter analysis object (in contrast to the analysis object Revolution Synchronous Order Tracking) allows the calculation and return of the filtered orders in the original time domain. The analysis object can therefore also be used for acoustic analyses of the filtered orders.

Two measurement methods are predominantly used for order tracking. During a ramp-up, the vibration and the instantaneous speed are measured synchronously, while the machine is usually ramped up slowly from its minimum to its maximum speed (run-up analysis). In a second method, the machine is first brought up to a certain speed and then a vibration measurement is taken for this speed (order tracking with constant speed). The Order Filter analysis object supports both variants. Order tracking using revolution synchronous resampling is also very flexible and can even be performed for noisy as well as non-monotonic speed data sets.order tracking with variable speed).

Data Tab

This is the tab where the input data and parameters for transformation from the time to revolution domain are specified (for details, also see RevolutionSyncSampling):.

Signals in the Time Domain

The time signal to be analyzed for the analysis object must be in the data structure Signal. The Speed can be specified either in the Signal data structure (corresponds to a run-up analysis with variable speed) or as a scalar value (corresponds to order tracking with constant speed or to a fixed fundamental frequency).

If there are no units or unit management is switched off, the speed is always interpreted in the unit [1/min] and the X component of the time signal is interpreted in the unit [s].

The instantaneous speed is often measured with a pulse encoder, which records a certain number of pulses per revolution. You can convert the resulting pulse signal directly into a speed signal. To do this, select the option Speed is pulse signal and specify the Number of pulses per revolution. The conversion of the pulse signal into a speed signal is done with the ImpulseToFrequency function.

Resampling into the Revolution Domain

Three different modes are available for resampling from the time to the revolution domain. In most practical applications, the linear resampling is sufficient:

If the splines or the FFT resampling method is selected, a resample factor must be specified by which the sampling rate of the signal is increased during the transformation algorithm.

Regardless of the resampling method selected, the Number of data points per revolution for the signal transformed into the revolution domain must be specified. This determines the sampling of the transformed signal in the revolution domain. According to the Nyquist sampling theorem half of the value determines the maximum order that can even be calculated.

To determine the Number of data points per revolution two modes are available:

Options Tab

In this tab the Order Filter is specified for bandpass filtering of single or multiple orders. Here, ordinary IIR bandpass filtering is performed around the respective order of the signal transformed into the revolution domain (for details, also see RevolutionSyncSampling).

Output

The following four output modes for the Order Filter analysis object are available:

Bandpass Filter

For calculating order curves, ordinary IIR bandpass filtering is performed around the respective order of the signal transformed into the revolution domain. The following bandpass filters are available:

Orders to Filter

The orders to be filtered can be explicitly set in a table or taken from a data set. The selected orders form the midband frequency of the set bandpass filter. Order bandwidths can be expressed symmetrically around the order as 3dB filter bandwidths. The 3dB filter bandwidth defines the order range, outside of which frequencies are blocked out (defined by -3 dB point, i.e. point at which 70.7 percent of the original signal amplitude is reached).

RMS Calculation

The following two RMS modes are available for calculating the RMS curves of the filtered orders (as a list or signal series) or for RMS calculation of the sum of the filtered orders:

Example of an order filter with block-wise RMS calculation for a run-up (with Y = amplitude, X = speed):

Result

For all output modes (with the exception of averaged RMS order tracking), the following result types are available:

For result Y = amplitude, X = speed, the result contains non-equidistant speed sampling points (these correspond exactly to the equidistant revolution sampling points after filtering in the revolution domain). The result can therefore also be evaluated across an equidistant speed range with the Speed start, Speed end and Speed increment to be defined. Here, linear interpolation is performed across the specified equidistant speed range. For linear resampling to provide meaningful results, the speed resolution calculated by the algorithm should be sufficiently high (e.g. moving RMS calculation or block-wise RMS calculation with sufficiently small block size).

Note The calculation of the (blockwise) RMS curves of the orders to be filtered provides curves similar to the analysis object Revolution Synchronous Order Tracking Analysis. For this purpose, in the latter analysis object the output mode Order cuts including RMS level extraction in order band must be set. The averaged RMS order analysis of both analysis objects provides the same results.

Example of an order filter across time for a run-up (Y = amplitude, X = time):

Example of averaged RMS order analysis for a constant speed signal (RMS of each filtered order):

Examples

In the project database C:\Users\Public\Documents\Weisang\FlexPro\2023\Examples\OrderT racking Analysis.fpd or C:>Users>Public>Public Documents>Weisang>FlexPro>2023\Examples\Order Tracking Analysis.fpd you will find examples of the different use cases and modes for which an order analysis can be performed. In particular, the diagrams above are included there.

FPScript Functions Used

RevolutionSyncSampling

IIRFilter

IIRPeakFilter

ImpulseToFrequency

See Also

Analysis Objects

Order Tracking Option

Revolution Synchronous Order Tracking Analysis Object

Revolution Synchronous Sampling Analysis Object

IIR Filter

Harmonic Filter