|
|
Contact Us
|
|
|
 |
At left...
Periodicity in rotating machinery is a key indicator of
mechanical processes with the mechanism. Time histories re-sampled
to revolution domain will guarantee a fixed number of data points
per revolution independent of speed. When an ensemble of contiguous
engine cycles is stacked and viewed top down, the resulting display
can reveal synchronous and non-synchronous signal components. |
Technology...
is a core distinction we bring to
our customers. We continuously strive to diversify and advance our technology
portfolio so that we can exceed our customer's expectations and lead
through innovation.
Automated Machine
Noise Recognition
|
 |
At Signal.X, we are developing technology for automated
classification of inherently non-stationary sensor data.
Time, frequency or time/frequency metrics are used to build
feature vectors that encode the instantaneous behavior of
signals. Feature strings can be used to train a recognition
engine that learns the time evolution of underlying
statistical process assigning states to common clusters
of features. The resulting state sequence is then a probabilistic
fingerprint of the process, and can be used to score similarity
to training data. The method has application for machine
tool health, laboratory test system monitoring, production
quality screening and any other situation where automated
recognition of complex non-stationary machine data is needed.
|
|
Machine tools state sequence detected automatically.
|
Rotating Machinery Diagnosis
The Short Time Fourier Transform (or
Gabor Transform) shows the interplay between resonance
at constant frequency and harmonic content proportional
to shaft speed
|
The complex dynamics in rotating
machinery are excited by forces generated both
mechanically and electrically. In many cases, the
dynamics of the driven load are equally as important as
the driver. Signal.X will apply a variety of standard
NVH analysis methods to better understand driveline
dynamics. For example, we have recently added torsional
vibration analysis and multi-plance balancing to our
MajX-DSA product to help round out the suit of
capabilities needed to properly evaluate, quantify and
correct anomalies in rotating machinery behavior.
|
Time/Frequency
Analysis Methods
A wide-open-throttle sweep with shifts
is easily tracked with Gabor Order Tracking
|
Signal.X has applied the Gabor transform and its inverse,
the Gabor expansion, as implemented by National Instruments
Corporation in the Order analysis Toolkit for LabVIEW, to
several unique applications. The method allows arbitrary
time/frequency filtering through masking and reconstruction
of desired signal components. Masks can be arbitrarily designed
and weighted allowing partial suppression of some components
and amplification of others. When adapted to order analysis,
masks are proportional to RPM data and allow reconstruction
of time domain order content at the native sample rate of
the data. The algorithm's key strength is the intuitive
time/frequency display and the extreme speed of the extraction.
(pdf) - Paper Presented at the SAE Noise
& Vibration Conference May 2001
|
Transient
Detection
FFT and model-based spectra (in red)
of a single impact event during gear rattle
|
Signal.X has developed an approach to automated
transient detection for use in production and laboratory audit
applications. Rattle impact generates short duration broadband
energy at a rapid repetition rate. The impact will
excite resonant response, which can be sensed with accelerometers
and microphones. Separation of adjacent impacts challenges FFT
based spectral analysis because of the need for very small block
sizes. Signal.X is addressing this problem with model based
spectral analysis methods. Super-resolution spectra can be defined
with just a small sample of data (i.e. 32 points sampled at
20 kHz) containing only a single impact event. Resonant response
of the structure can be clearly identified and used to discriminate
impulse types and intensity. The technique has led to a fast
and effective rattle metric suitable for automated test applications.
|
Machine
Vision
Triggered acquisition of gear contact patterns
freeze the image in the frame allowing accurate metric definition
|
A comprehensive process control strategy
linking gear contact patterns with end-of-line NVH performance
is the motivation for automating contact pattern analysis.
Triggered vision systems produce highly accurate image identification
in the form of feature vectors assigned to each tooth. Proprietary
Signal.X recognition algorithms can then be taught to classify
patterns in the feature data in terms of causal factors in the
assembly.
|
|
 |
|
