Shield will support two native sample rates that are selected by DSA device type. The fast group is for dynamic channels such as microphones and accelerometers. A slow group will accommodate support data conditioned to a voltage. CAN bus data can be included and will be re-sampled to match the system’s slow sample rate group. Counters enable the acquisition of TTL pulse trains proportional to shaft speed. These can be converted to RPM vs time or torsion (degrees per second) vs time and are presented at the sample rate of the fast group.
Recommended 4-, 8- and 16-channel DSA devices from National Instruments:
There are many DSA devices available from National Instruments that can provide voltage measurement for the slow channel group. We commonly use the NI PXI 6281 and 6284, which offer high dynamic range, wide sample rate capabilities and 8 or 16 channels of differential input. Note that devices in the slow channel group and the fast channel group are triggered to begin acquisition simultaneously. However, they operate on separate clocks and the groups are not synchronized.
Recommended DSA devices for signals placed in the slow channel group:
Shield will support single- or dual-port CAN devices.
NI-XNET is integrated within Shield to permit user-supplied database definition files (.ncd, .dbc) to be created, edited and deployed. Channels acquired through CAN are re-sampled to a rate consistent with the slow sample rate group and appear as additional channels on the channel list.
Shield will support counters to detect TTL level pulse trains proportional to shaft speed. Up to 4 RPM signals can be captured for use in computations such as order extraction and revolution domain conversion (adaptive re-sampling). In addition, counter inputs can be further processed to RPM vs time and torsional vibration (degrees per second) vs time. These will appear as additional dynamic channels in the fast channel group.
Pre-processing capabilities are utilized to modify existing dynamic signals and to create new virtual signals through math manipulation of existing signals. These operations are enacted on data read from a data file and prior to presenting them for further analysis in Shield.
Modifications capabilities include:
- Single and double integration and differentiation (for signals defined with MKS units)
- Conversion to inches per second and mils for vibration signals define in G’s
- Low-pass, high-pass, band-pass and band-stop filtering
- Sample rate conversion (unique rate per channel may be defined)
- Acoustic and human-body weighting (i.e. A-weighting for microphone signals)
- Conversion to signal envelope
- Virtual channel synthesis editor (create virtual channels from algebraic combinations of existing channels)
- Counter inputs can be further processed to RPM vs time and torsional vibration (degrees per second) vs time
In Shield, triggers act on acquired data to locate spots in the file. Triggers are defined through a parameter set that controls a look-up algorithm. The result of trigger detection is an index pointing to a spot in the file where the specific trigger condition is met. Triggers can be defined on any channel. For example, when a speed is reached, or when the torque reaches a threshold during a torque ramp.
An order relative to a shaft rotation is defined as the periodic response to periodic excitation modified by path dynamics. Order energy extraction is accomplished with the Gabor Order Tracking method. An RPM proportional mask is used to select order-related energy in the time/frequency domain and an inverse transformation is used to recover this energy in the form of a time history. From this time history, an RMS envelope is determined as the order magnitude vs time.
RMS magnitude, vector and peak hold averaging are available to produce power spectra of dynamic channels using linear and exponential weighting modes. The method of cepstral analysis is provided to allow the definition of a lifted power spectrum. 1/1. 1/3. 1/12. 1/24th octave spectral analysis is provided to comply with the IEC 1260 standard.
For periodic transient detection, a proprietary methodology is provided that breaks a time or revolution domain signal into an ensemble of synchronous frames that will reinforce periodic events. Conditioning formulations act on the ensemble to produce an impulse indicator waveform. For randomly occurring transients, a combination of band-pass filtering and enveloping of dynamic signals will produce an impulse indicator waveform.
Time/Angle Waveform Processing
All time or revolution domain signals can be acted on directly by a variety of operators to achieve waveforms from which metrics can be derived. These operators will act on user-defined overlapping frames, which reduce the resulting waveform density to the frame count and enhance the extraction of metrics.
Signal operator capabilities include:
- Crest factor
Features to Metrics
Waveforms such as orders, spectra, impulse indicators and other modified signals are subject to a variety of discrete metric definition options.
Examples include mean, median, interquartile (25th – 75th percentile), etc.
Examples include max, min, range, location of max/min/range, etc.
Peaks above a threshold are placed in a peak extraction list. The list is further operated upon to reduce it to metrics such as nth highest peak, max peak, mean of peaks, etc.
Waveforms are compared to upper, lower or tolerance functional limits to define an out-of-limit list (OOL). Exceedance operators such as mean of OOL values, max of OOL value, etc. reduce the list to a single number metric.
Pass/Fail Decision Making
Individual metrics are tested against limits, bounded limits or each other to define pass/fail. Each row of a decision table represents an independent metric formulation and the pass/fail results of each are ordered together to define overall test pass/fail.
Metrics can be used in combination with other metrics in structured nth order polynomial form to make new metrics. A freeform equation editor permits metrics to become parameters in a hierarchical formula that is suited to defining arbitrarily complex computations such as efficiency.
Boolean Logical Metrics
Numeric metrics can be converted to Boolean metrics (TRUE or FALSE) by comparison to numeric limits, numeric limit ranges and in logical operations with other numeric metrics. Boolean state can be used as pass/fail criteria. Nested Boolean formulas can also be defined using Boolean metrics to create more complex pass/fail behaviors.
Shield has a file-import/export utility that supports the following file types: .tdms, .mat, .wav, .uff, .csv, .txt, .hdf. The utility facilitates import of each of these formats to .tdms for use in Shield. Export from .tdms to all formats except .mat is supported. Note that some formats will not support multiple sample rate groups and only the fast-channel group may be exported.