Analytical system and method for testing gears

Through rotary sampling and model basis function technology, the dynamic behavior of the gear set is analyzed and displayed, and the problem of difficult to effectively analyze and display the dynamic behavior of the gear set in the prior art is solved, and a deeper understanding of gear set performance and real-time monitoring are achieved.

CN114424041BActive Publication Date: 2024-11-15THE GLEASON WORKS
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202080064789.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2019-10-09
Filing Date
2020-10-08
Publication Date
2024-11-15
Estimated Expiration
2040-10-08

AI Technical Summary

Technical Problem

Existing gear set testing technologies are difficult to effectively analyze and display the dynamic behavior and mass characteristics of gear sets, especially motion transmission errors, noise and vibration characteristics, and the processing and analysis of sensor outputs lacks flexibility and practicality.

Method used

The sensor output signal is processed using rotary sampling technology, and the model is projected and reconstructed through the model basis function to identify the quality characteristics of the gear set, and provides dynamic analysis and visualization methods, allowing users to interactively adjust the analysis parameters to achieve autonomous and timely data processing and Anomaly detection.

Benefits of technology

The ability to extract dynamic behavior hidden in static data provides a deeper understanding of gear set performance, improves gear set testing flexibility and practicality, and enables real-time monitoring and identification of gear set quality characteristics and potential problems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114424041B_ABST
    Figure CN114424041B_ABST
Patent Text Reader

Abstract

A gear rolling test method involving the analysis and display of gear set performance data, including kinematic transmission errors, acquired on a gear set rolling tester or gearbox test fixture, and including analysis and visualization of this data in the time domain, frequency domain, and a mixture thereof. The invention also includes: fundamental improvements to core signal processing, analysis, and signal processing sequences that allow the data to be explored in a more insightful manner; methods for visualizing and reporting the data and results; and a human-machine user interface paradigm for providing these functions and features with greater flexibility and utility.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the processing, analysis and display of sensor outputs, particularly with respect to the testing of gears. Background Art

[0002] The present invention relates generally to the processing, analysis and display of various sensor outputs, and particularly to sensor outputs related to gear set motion transmission error (MTE), noise and vibration phenomena that can be generated by rolling test gear sets in test machines such as those shown and described in U.S. Patents Nos. 6,120,355 and 7,553,115, the disclosures of which are incorporated herein by reference. The present invention may also be performed on a functional test platform for rolling test gears, such as a single flank tester or a dual flank tester, in which the relative movement between the axes of the gears is measured as the gear pairs rotate in meshing engagement, such as described in US 2020 / 0292305. Figure 1 The disclosure of which is incorporated herein by reference.

[0003] Sensors that produce outputs to which these methods can be applied include rotational position measurement devices such as encoders and resolvers; vibration measurement devices such as single-axis and multi-axis accelerometers and acoustic sensors; and other motion or displacement measurement devices such as optical glass scales, magnetic scales, optical, laser and ultrasonic devices, and capacitive or inductive displacement sensors.

[0004] The sensors can be arranged in many configurations and arrangements. Rotational sensors typically monitor two main shafts, in which the driving and driven members of the gear set are mounted. In cases where the device under test includes more than just a single gear pair (e.g., including a gearbox or planetary assembly), additional rotary encoders can measure the motion of other rotating elements.

[0005] Linear, vibration, acoustic and other sensors can be mounted in fixed locations on or near the machine, or mounted on linear or rotational moving elements of the machine (e.g., oscillating heads and linear axes). For example, a displacement measuring device can be applied to a double-toothed gear set tester, and the output can also be analyzed by the system of the present invention. Vibration sensors can respond to structure-borne vibrations, air-borne vibrations, or a combination of both.

[0006] One or more accelerometers can also be mounted on the rotating spindle to measure rotational motion, including acceleration, velocity, and position.

[0007] As the mounting components of the gear set roll under given speed and load conditions determined by the testing machine, vibrations and other movements are induced depending on the nature and mass characteristics of the gear set being tested. Summary of the invention

[0008] It is an object of the present invention to use the aforementioned testing machine sensors to provide data and signals that can be processed and analyzed to provide insight into the operating and quality characteristics of a gear set, particularly the error motion, noise and vibration characteristics.

[0009] A further object of the present invention is to process and analyze mathematically derived data and signals to provide additional understanding of the operation and quality characteristics of the gear set. For example, these data and signals can be obtained from computer simulations or calculated products of sensor data, such as acceleration determined from analysis of position and time information.

[0010] Data or signals comprising inputs to the system of the present invention can be sampled and understood with reference to time (temporal sampling) or with reference to the rotational motion of one or more of the rotating gear members (rotational sampling). In the analysis of the aforementioned sensor signals associated with the motion of the gear set, it is often of particular value to represent the signals in terms of rotational sampling. When the present invention employs rotational sampling, time information is saved to facilitate conversion between time and rotational sampling when necessary.

[0011] The present invention relates to the analysis and display of gear set performance data, including kinematic transmission errors, acquired on a gear set rolling tester or gearbox test fixture, and includes analysis and visualization of such data in the time domain, frequency domain, and a mixture thereof. The present invention also includes: fundamental improvements to core signal processing, analysis, and signal processing sequences that allow data to be explored in a more insightful manner; methods for visualizing and reporting data and results; and a human-machine user interface paradigm for providing these functions and features with greater flexibility and utility.

[0012] The outstanding aspect of the method of the present invention is the ability to extract and display dynamic behavior that has heretofore been hidden within data records that have been viewed as essentially static. Other aspects include the ability to track or attribute data and results to individual teeth on the gear set member under test ("indexing" of the teeth) and the ability for the user to interact with the display to dynamically change the parameters of both analysis and visualization. In addition to conventional "contract testing" (post-processing of the collected data records according to strictly predefined test and analysis parameters), these functions and features are provided for the gear set rolling tester in a new flexible "always on" paradigm, whereby data collection, processing, analysis, anomaly detection, etc. are performed autonomously and timely whenever possible without waiting for specific instructions from a human user or waiting for a predefined cycle to be started. In either method of testing and analysis, the human user is also able to adjust the parameters for analysis and display in an exploratory and interactive manner. In the present invention, it should be understood that the terms "user" and "operator" are equivalent.

[0013] A first embodiment of the present invention is directed to a method of testing a gear, comprising providing a gear pair comprising a first gear and a second gear, wherein the first gear and the second gear are operable to roll in meshing engagement with each other. The first gear rolls with the second gear on a gear testing machine, wherein the gear testing machine is configured with one or more sensors operable to provide output signals indicative of rolling characteristics of the one or more gears. At least one independent signal is generated, wherein the independent signal is at least one of a signal based on an output from one or more of the sensors or an analytically derived signal. At least one frame portion having a non-zero duration is selected from the at least one independent signal, wherein each of the selected frame portions is processed according to the following steps:

[0014] determining a set of model basis functions based on the selected frame portion,

[0015] Projecting the selected frame portion onto the model basis functions to produce model coefficients,

[0016] Determine the reconstruction model from the model coefficients and model basis functions,

[0017] The reconstructed model is subtracted from the selected portion of the frame to produce a residual,

[0018] The method further comprises at least one of the following:

[0019] (a) displaying and analyzing the residuals to identify a quality characteristic of at least one of the first gear and the second gear, and accepting or rejecting at least one of the first gear and the second gear based on the quality characteristic, and,

[0020] (b) displaying and analyzing the model coefficients or reconstructing the model to identify quality characteristics of at least one of the first gear and the second gear, and accepting or rejecting the at least one of the first gear and the second gear based on the quality characteristics.

[0021] In a preferred embodiment, the model basis function is determined from at least one of: (a) the duration of a selected frame portion, (b) the gear ratio of the first gear to the second gear, (c) the fundamental harmonic frequency generated by the rotation of the first gear and the second gear or the duration of the meshing period of the first gear and the second gear, (d) a set of predetermined harmonics of each fundamental frequency, the set including zero, and (e) a set of non-periodic basis functions defined within the duration of the selected frame portion, the set including zero.

[0022] In another preferred embodiment, the reconstructed model and residuals are continuously re-evaluated as additional information is obtained from sensors during machine operation.

[0023] In another preferred embodiment, multiple independent signals are sampled simultaneously at consistent intervals to facilitate direct correlation and comparison.

[0024] In yet another preferred embodiment, at least one independent signal is collected at uniform time samples and subsequently interpolated to produce a sequence of samples regularly spaced in the position of one of the first gear and the second gear.

[0025] In further embodiments, the display and analysis results are retained on an electronic storage medium for recall, wherein the display and comparison of more than one previous record is accomplished by one or more of the following:

[0026] (a) Simultaneous adjacent presentation,

[0027] (b) sequential presentation with progression guided by user interaction, and

[0028] (c) Sequential presentation with automatic, animated progression.

[0029] The present invention additionally discloses a display and method for displaying residuals, reconstruction models or a combination thereof, which deserves independent protection, and the method comprises:

[0030] (a) subdividing the selected frame portion into a plurality of contiguous signal segments of equal duration, wherein the duration is equal to one or more of the following:

[0031] 1. One or more complete meshing cycles,

[0032] 2. One or more complete revolutions of any gear in the gear set,

[0033] (b) representing the signal segment as a function of one or more of:

[0034] 1. Time,

[0035] 2. The position of one of the gears in the gear set,

[0036] 3. Frequency, with the help of Fourier transform,

[0037] (c) presenting the signal segment functions aligned and overlaid on the same independent axis as one or more of the following:

[0038] 1. The trace of the signal fragment function,

[0039] 2. The trace of the statistical variation of the signal segment function, consisting of the arithmetic mean and / or variance, determined point by point.

[0040] The present invention also discloses a display and method for displaying residuals, reconstruction models or a combination thereof, which deserves independent protection, and the method comprises:

[0041] (a) subdividing the selected frame portion into a plurality of contiguous signal segments of equal duration, wherein the duration is equal to one or more complete meshing cycles,

[0042] (b) determining for each signal segment a scalar metric for each signal segment consisting of one or more of the following:

[0043] 1. Peak-to-peak variation,

[0044] 2. RMS amplitude,

[0045] 3. the temporal or rotational shift of the signal segment required to optimally align the signal segment with the reference, which is determined as either: the average of at least two segments or the average of a subset of the reconstructed models having the same duration as the signal segment,

[0046] 4. A measure of similarity to a reference, determined as either: the average of at least two segments or the average of a subset of the reconstructed models having the same duration as the signal segments, the measure consisting of one or more of: Pearson correlation coefficient and / or root mean square error,

[0047] (c) Each fragment is assigned two indices, the first index corresponding to any of the following:

[0048] 1. The number of complete revolutions of one of the first gear or the second gear observed in the segment signal before the segment duration begins, or

[0049] 2. The number of engagement cycles observed before the segment duration begins,

[0050] modulo the number of teeth on one of the first gear or the second gear; the second index corresponds to any of:

[0051] 3. the number of complete revolutions of the other of the first gear or the second gear observed in the segment signal before the segment duration begins, or

[0052] 4. The number of meshing cycles observed before the segment duration begins, modulo the number of teeth on the other of the first gear or the second gear,

[0053] (d) presenting a scalar metric mapped to a series of gray levels or color values ​​onto a two-dimensional image, where one coordinate of the image corresponds to a first index and another coordinate of the image corresponds to a second index.

[0054] The present invention additionally discloses a display and method worthy of independent protection, wherein displaying the residual, the reconstruction model or a combination thereof comprises:

[0055] (a) subdividing the selected frame portion into a plurality of contiguous or overlapping signal segments of predetermined equal duration, each signal segment being indexed sequentially,

[0056] (b) Representing the signal fragment as a spectral function with the aid of Fourier transform,

[0057] (c) determining from the spectral function one or more of the following measures:

[0058] 1. Amplitude,

[0059] 2. Logarithmic amplitude,

[0060] 3. Phase,

[0061] 4. Phase drift, which is the difference between the actual phase and the phase predicted by advancing the phase of the previous spectral function by the time or rotation interval between the two segments,

[0062] (d) Presenting the spectral measure with values ​​mapped to a series of gray levels or color values ​​as a two-dimensional image, where one coordinate of the image corresponds to the index of the segment and another coordinate of the image corresponds to the frequency variable of the spectral function.

[0063] The present invention also discloses a display and method worthy of independent protection that displays traces of one or more of residuals, reconstructed models, or a combination thereof on a graph, wherein the independent axis represents the time or position of one of the first gear or the second gear in the gear set, and the subordinate axis represents the amplitude of the displayed quantity.

[0064] The present invention additionally discloses a display and method for displaying model coefficients worthy of independent protection, the method comprising plotting one or more of the following:

[0065] (a) a bar graph with one bar per coefficient and height proportional to the magnitude of the coefficient; and

[0066] (b) one or more radial graphs on which are drawn line segments, one for each coefficient, radiating from the coordinate origin and having a length proportional to the complex magnitude of the corresponding coefficient and an angle with the horizontal axis proportional to the complex phase of the corresponding coefficient.

[0067] The present invention additionally discloses a method of testing a gear worthy of independent protection, the method comprising providing a gear pair including a first gear and a second gear, wherein the first gear and the second gear are operable to roll in meshing engagement with each other. The first gear rolls together with the second gear on a gear testing machine, wherein the gear testing machine is configured with one or more sensors operable to provide output signals indicative of rolling characteristics of one or more gears. At least one independent signal is generated, wherein the independent signal is at least one of a signal based on an output from one or more of the sensors or an analytically derived signal. At least one frame portion having a non-zero duration is selected from the at least one independent signal, wherein each of the selected frame portions is processed according to the following steps:

[0068] determining a set of model basis functions based on the selected frame portion,

[0069] Projecting the selected frame portion onto the model basis functions to produce model coefficients,

[0070] Determine the reconstruction model from the model coefficients and model basis functions,

[0071] The reconstructed model is subtracted from the selected portion of the frame to produce a residual,

[0072] The method further comprises at least one of the following:

[0073] (a) displaying and analyzing the residuals to identify a quality characteristic of at least one of the first gear and the second gear, and accepting or rejecting at least one of the first gear and the second gear based on the quality characteristic, and,

[0074] (b) displaying and analyzing the model coefficients or reconstructing the model to identify quality characteristics of at least one of the first gear and the second gear, and accepting or rejecting the at least one of the first gear and the second gear based on the quality characteristics,

[0075] Wherein the method is performed in an always-on mode, wherein the previous steps are repeated and continuously performed as long as the one or more sensors detect coordinated movement between the gear pairs and a new data frame can be filled.

[0076] In a preferred embodiment, in the always-on mode, the gear ratio of the first and second gears being scrolled is not known in advance, but the always-on mode determines the gear ratio based on real-time data it receives.

[0077] In a further embodiment, a user is able to view information comprising at least one of a measurement, a display, a visualization, and a result at any time while the always-on mode is operating, wherein the information is available regardless of whether any other test mode is simultaneously active.

[0078] In yet another embodiment, the information comprising at least one of a measurement, a display, a visualization, and a result comprises at least one of:

[0079] (a) the amplitude and phase of one or more harmonics of a frequency associated with the meshing period;

[0080] (b) Average signal shape over the duration of one meshing cycle;

[0081] (c) gear runout harmonic amplitude;

[0082] (d) Average signal shape of gear runout;

[0083] (e) the presence and location of tooth lesions;

[0084] (f) removing the average signal shape of the pinion base; and

[0085] (g) Average signal shape after removing the gear base.

[0086] In yet another embodiment, the method further comprises interactively manipulating the display to better examine aspects of the displayed data.

[0087] In another embodiment of the present invention, the display can be paused so that the current set of displayed data can be examined without being replaced by a subsequent frame.

[0088] In a further embodiment of the present invention, a current set of data can be captured and stored for later retrieval or verification and / or for comparison with current data.

[0089] Additional embodiments of the invention disclose that differences between current data and stored reference data or master data are checked and visualized.

[0090] In yet another embodiment of the present invention, the always-on mode further includes identifying and locating tooth damage on at least one of the first gear and the second gear without separately establishing any test parameters or settings, the always-on mode identifying the component rotation or tooth index that best represents the location of the identified damage, and the always-on mode automatically notifies the user of such damage via a text message and / or a graphical display. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] Figure 1 A graph depicting always-on monitoring of encoder counts and MTE calculations is shown.

[0092] Figure 2 A preliminary analysis of streaming MTE data in a specific channel is depicted.

[0093] Figure 3 A component diagram depicting the stages of the harmonic decomposition process is shown.

[0094] Figure 4 Depicted is the display of the MTE data decomposed into gear, pinion, and mesh component bases, Legendre polynomial bases, and residuals for subsequent analysis.

[0095] Figure 5 A display of the magnitudes of the decomposed gear, pinion, and mesh component bases of example MTE data is depicted.

[0096] Figure 6 A discrete Fourier transform view of the MTE data is shown, where harmonics of the gear, pinion, and mesh frequencies are highlighted, and frequencies not related to the gear, pinion, or mesh harmonics are deemphasized.

[0097] Figure 7 A display of an overlay of multiple MTE data frames at the meshing cycle level on a common axis is shown without model subtraction that would compensate for low frequency variations between frames.

[0098] Figure 8 Draws Figure 7 Display of the same overlay depicted in , but with the modeled gear and pinion harmonic components subtracted from the signal to highlight the MTE variations that are directly attributable to the individual tooth contacts between the components. The mean of the overlay frame is depicted in black, and the variance bands are represented by the shaded grey areas surrounded by dashed lines, both superimposed on the individual frame plots.

[0099] Fig. 9 Depicted is a display of multiple MTE data frames at the gear cycle level, superimposed on a common axis and with the modeled pinion level and mesh level components subtracted from each frame to emphasize only the portion of the MTE that is directly attributable to the gear change. The average MTE is shown in the black superimposition.

[0100] Fig.10 Plotted is a display of many MTE data frames at the pinion period level, superimposed on a common axis and with the modeled gear stage components subtracted from each frame to emphasize only the portion of the MTE attributable to the pinion and individual tooth variations. The average MTE over the displayed frames is depicted in a black overlay.

[0101] Fig.11 The position coverage diagram (similar to Figure 8 ) is a display of multiple spectral amplitudes obtained by discrete Fourier transform of each frame in the MTE. The average amplitude (computed in the amplitude sense only) as a function of frequency over all MTE frames is depicted by the black bars superimposed on the grayscale distribution of the individual spectral amplitudes.

[0102] Fig.12 Shows Fig.12The same overlay of spectral amplitudes depicted in , but the horizontal axis is shifted to show a different frequency range and the average amplitude is evaluated in a complex phase sense.

[0103] Fig.13 Plotted is a display of a spectrogram of the MTE data with the gear stage and pinion stage components subtracted. The position and frequency content are displayed in units of mesh period and fundamental mesh harmonic, respectively, with the grayscale value indicating the spectral amplitude at a given position and frequency bin.

[0104] Fig.14 Draws from Fig.13 The calculated periodogram of the spectrogram shown is shown, along with the standard deviation of the spectral amplitude at each frequency and the total RMS amplitude calculated over the bandwidth of a single meshing harmonic.

[0105] Fig.15 A display of statistical measures calculated over successive mesh cycle frames of MTE data is shown, the statistical measures being mapped to individual contact cycles between specific gear and pinion teeth and represented by grayscale or color values.

[0106] Fig.16 Depicted for supplementation Fig.15 A grid view of the MTE data for a specific mesh cycle frame along with a reference curve used as the average per-mesh MTE in the evaluation of some per-mesh statistical measures.

[0107] Fig.17 Depicted is the separation of an input signal (eg, the output of a sensor) into data frames of specific duration suitable for subsequent display and analysis. Fig.17 Also shown is the separation of frames into segments suitable for subsequent display and analysis.

[0108] Fig.18 An example of multiple channels that can be employed in the present invention is shown. DETAILED DESCRIPTION

[0109] The terms "invention", "the invention" and "the present invention" as used in this specification are intended to refer broadly to all subject matter of this specification and any patent claims below. Statements containing these terms should not be understood to limit the subject matter described herein or to limit the meaning or scope of any patent claim below. In addition, this specification does not seek to describe or limit the subject matter covered by any claim in any particular part, paragraph, statement or figure of the application. The subject matter should be understood by reference to the entire specification, all drawings and any claims attached. The present invention can use other configurations and can be practiced or implemented in various ways. Moreover, it should be understood that the words and terms used herein are for descriptive purposes and should not be considered limiting.

[0110] As used herein, "include", "have" and "comprises" and variations thereof are intended to cover the items listed thereafter and their equivalents as well as additional items. The use of letters to identify elements of a method or process is for identification purposes only and is not meant to indicate that the elements should be performed in a particular order. In addition, terms such as "first", "second", "third", etc. are used herein for descriptive purposes and are not intended to indicate or imply importance or significance. As used herein, the singular forms "a", "an", and "the" are also intended to include plural forms unless the context clearly indicates otherwise. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items and may be abbreviated as " / ".

[0111] The details of the present invention will now be discussed with reference to the accompanying drawings, which illustrate the present invention only by way of example. In the accompanying drawings, similar features or components will be referred to by similar reference numerals. Where applicable, various embodiments can include and / or exclude different aspects, features and / or advantages. In addition, where applicable, various embodiments can combine one or more aspects or features of other embodiments. The description of the aspects, features and / or advantages of a specific embodiment should not be interpreted as limiting other embodiments or claims.

[0112] The present invention includes analyzing any disturbance associated with the rotational motion of two or more gears. Currently, rotational motion transmission errors (i.e., the difference between the actual motion of the output gear and the ideal motion as a function of the actual motion of the input gear), linear displacement in conventional double-flank testing, and vibration acceleration of the test structure can be considered. The present invention can be performed on a computer that is a component of the testing machine or on a computer remote from the testing machine. The information obtained from the analysis of the present invention can be sent directly to a gear processing machine (e.g., a cutting machine or a grinder) via a closed-loop system for continuous process improvement. The information obtained can be stored for subsequent research and analysis. It should be noted that in the present invention, any member of the gear set can be identified as an "input" gear, and therefore, the other member will be identified as an "output" gear. However, in practice, the member with fewer teeth is typically identified as the input gear.

[0113] The method of the present invention continuously monitors machine sensors, such as those of the type described above, to determine when two or more gears in a test apparatus are in coordinated motion. Motion may be automatically indicated by a non-zero correlation between rotary encoders on the input or output gears or by the crossing of a predetermined threshold of vibration acceleration at a frequency predicted by the gear set configuration.

[0114] Whenever coordinated motion is detected, data is recorded for further analysis. The rotational position of the input and output gears is measured via rotary encoders attached to the input and output spindles. Vibration acceleration is measured via one or more accelerometers attached to the test equipment. All measurements are time sampled, where sensor readings are captured at regular time intervals.

[0115] The rotational positions of the input and output gears in the gear set are interpolated (e.g., by local or global polynomial methods) and resampled along with any additional time sampled data so that all measured samples occur at consistent times and the position of the input gear advances at uniform intervals (referred to herein as rotational samples). Irregularly spaced time values ​​(which correspond to specific moments of the resampled time data) are retained as additional measurements that capture information about the acceleration of the input gear. All consistent measurements of a given position of the input gear represent a measurement "packet" of captured data for further analysis.

[0116] By subtracting the ideal position predicted by the gear ratio and the coincident position of the input gear from the output gear position, the rotational position of the output gear is converted into a motion transfer error measure.

[0117] From the rotational sampled positions of the input and output gears, the effective gear ratio of the gear set is determined by linear regression of the observed motion. The effective ratio is converted into an estimate of the number of teeth on the input and output gears by assuming that the gear set consists of two gears and converting the floating point ratio into a rational expression, where the number of teeth on the input gear is the numerator and the number of teeth on the output gear is the denominator.

[0118] Conversion of the floating point ratio to a rational expression can be accomplished by limiting the denominator to some maximum integer based on the application of the gear being tested, or by comparing the ratio to a preselected list of possible ratios to find the closest match.

[0119] A collection of samples from a single signal or a collection of samples as a "packet" of consistent samples from multiple signals can be accumulated into a frame (see Fig.17 , top). Each data frame represents all or a portion of the input signal starting at a desired starting sample and continuing for a predetermined duration. Multiple frames may overlap in duration (see Fig.17 , “overlapping frames”), or they can be disjoint; disjoint frames need not be contiguous in duration.

[0120] In conventional gear set testing, a test record or a defined series of records is collected, and a different analysis or DFT transform is then performed on each record. Therefore, these analyses only give results and conclusions about the average conditions within the record, but do not attempt to evaluate the variability of the characteristics within the record.

[0121] For example, when setting up a routine test, a record with a duration of 4 gear revolutions may be selected from which time and frequency domain results may be derived and displayed. An MTE measurement taken at a frequency of once per mesh may indicate an amplitude of 55 microradians, but this is essentially the average of this component measured over the length of the 4 gear revolution record. A closer examination of the data may show that the amplitude for the first two gear revolutions was 65 microradians, and the amplitude for the last two gear revolutions was 45 microradians. Exposing this amplitude variability may provide important clues about the quality characteristics of the gear set or the test conditions that are not revealed by conventional analysis methods.

[0122] On the other hand, in the test method of the present invention, multiple shorter frames can be defined within a particular test record. Analysis of one or more signals within each frame provides information about the more localized dynamic behavior of the gear set. Because frame overlap is allowed and the duration is arbitrary, in addition to the dynamic analysis facilitated by the method of the present invention, frames with a duration equal to the entire test record can also be defined to replicate current conventional test results.

[0123] The length of the frame can be chosen to correspond to various interesting phenomena associated with the motion of the gear set. In addition, collections of frames can be accumulated into "channels" that characterize sequences of related frames (see Fig.18 ). Fig.18 Examples of multiple channels that can be used with the present invention are depicted for an example 5:19 ratio gear pair. The track channel shown consists of frames that are one full track (5 gear revolutions) in length. The gear channel shown consists of a series of frames that are one full gear revolution in length. Three types of pinion channels are shown: Type A consists of a series of connected frames that are one full pinion revolution in length; Type B consists of a series of connected frames that are 4 pinion revolutions in length; and Type C consists of a series of overlapping frames that are 4 pinion revolutions in length. Finally, the meshing channel shown consists of a series of connected frames that are one meshing cycle in length.

[0124] For example, a channel may be defined as a set of one or more frames, each frame having a duration equal to an integer of any of the following:

[0125] A mesh period, which is the duration for which a single pair of teeth between any two gears in a gear set is expected to maintain a dominant effect on the relative motion of the components; the ideal mesh period duration can be obtained mathematically by dividing a complete rotation period of any gear by its number of teeth.

[0126] · One revolution of any gear in the gear set, or

[0127] • A "track" commonly referred to as a "chasing tooth" which is the least common multiple of the number of meshing cycles comprising a single revolution of each of any two interconnected gears in the gear set.

[0128] like Fig.18 As shown in the example of , the channel or frame duration may depend on the number of meshing cycles (teeth) that contain a single revolution of each member. If the user does not specify a gear ratio, an automatically determined estimate is used to identify the duration assuming a two-gear set. If the user specifies a gear ratio, perhaps involving more than two gears in a gear set, the channel duration is derived from the user-specified value.

[0129] Within each channel and frame, a core model of one or more collected measurements can be derived as a sum of one or more truncated Fourier series, each being an orthogonal basis and referred to herein as a "component basis". The fundamental frequency of each series is the frequency of rotation of one gear in the gear set expressed relative to one revolution of the input gear. Additionally, a component basis has a fundamental frequency corresponding to the fundamental frequency of a single meshing period, also expressed relative to one revolution of the input gear. In a gear set containing more than two gears, there may be more than one basic meshing period, in which case a component basis may be assigned to each meshing period.

[0130] If such a specification is made, the fundamental frequency of each component basis is determined based on the gear ratio and number of teeth specified by the user. Otherwise, the frequency is determined based on the number of teeth automatically estimated from the gear ratio determined assuming a two-gear set.

[0131] The cutoff for each component basis may be determined automatically, such as by an algorithm, or manually by user selection.

[0132] Within each frame, any component basis with a fundamental period longer than the duration of the frame is either excluded from the model entirely, or is altered so that only those components with a period shorter than the frame duration are represented in the model. The model for any frame can be supplemented by estimates of any long-term, excluded components determined in alternate channels of frames with sufficiently long duration.

[0133] The Fourier model may be augmented with one or more additional bases representing periodic phenomena external to the gear set under test, collectively referred to herein as "external bases," or with orthogonal basis functions, referred to herein as "non-periodic bases," representing non-periodic phenomena within the duration of each frame. A preferred embodiment of the present invention uses a truncated series of Legendre polynomials as the non-periodic basis to model behavior at levels longer than the maximum harmonic period.

[0134] The coefficients of all Fourier bases of the model can be determined simultaneously by fast Fourier methods, such as common fast Fourier transform (FFT). This is a conventional choice for finding the frequency domain harmonic content of such signals, and the disclosed system can provide such a solution. However, whenever there is non-periodic signal content, applying these known and efficient fast Fourier methods can lead to errors in identifying harmonic content. Therefore, the system of the present invention also includes an alternative method that can reduce or eliminate such errors.

[0135] One such alternative approach provided by the system simultaneously determines the coefficients of all bases of the model (including non-periodic or external bases) by least squares optimization, with arbitrary weights assigned to different bases to emphasize one base relative to another and resolve ambiguities caused by potential non-orthogonality of the different bases. This approach is less sensitive to "spectral leakage" errors associated with non-periodic content that plagues conventional fast Fourier analysis (e.g., when the frame to be analyzed contains a non-integer number of revolutions of one or more gears).

[0136] Another alternative method provided by the system iteratively determines the coefficients of the model by projecting the data, minus any previously determined model, onto each basis in turn in a predetermined but arbitrary order. The iterative process is repeated until the coefficients of any basis deviate by no more than a predetermined amount from the coefficients determined in the previous iteration. This method can be implemented by employing, for example, the well-known FFT, where spectral leakage is iteratively removed from the model. Note that in the general case of a completely unknown signal, the errors associated with FFT leakage are inherent and cannot be eliminated by any means. However, where it is known in advance that certain frequencies exist (e.g., given gear and pinion harmonics), the iterative projection technique can be "seeded" with these known periods, and the iterative projection method employed has great benefits.

[0137] After determining the core model in each frame for one or more measurements, the residuals for that frame and measurement are calculated by subtracting the model from the measurement. The residuals represent the portion of the measurement that cannot be explained by the model and provide a basis for further analysis of erratic gear set behavior.

[0138] When recording coordinated motion data, perform predetermined inference analysis on each channel, including but not limited to:

[0139] The instantaneous values ​​of the Fourier and other basis coefficients in each frame of the channel;

[0140] The instantaneous peak, RMS and average amplitudes over the duration of each channel;

[0141] The instantaneous "spread" of each Fourier basis coefficient, as the ratio of the power in the corresponding frequency to the total power in a spectral band centered at that frequency, the width of which is narrow enough so that it does not encroach on similar spectral bands around nearby components within the same Fourier basis.

[0142] In addition to the instantaneous values, the statistical variation of the calculated values ​​within the frame is calculated and reported for each channel.

[0143] Embodiments of the present invention encompass a measurement paradigm that includes an "always on" aspect and a "contract" aspect. The contract function is a conventional approach that has been known for many years and includes a test that is configured and defined in advance by the user according to various settings and parameters. It defines the test conditions, analysis, and reports to be generated throughout the test. This is the way in which gear testing machines with advanced measurement capabilities are currently considered to operate. In other words, little information is presented to the user other than that specifically requested and configured in advance by establishing a recipe (or "profile") for the test. If the position, motion, speed, load, or timing exhibited by the machine does not match the expectations (or "contract") of the monitoring system, a warning or alarm result is typically generated and the test is aborted.

[0144] In contrast, the always-on mode of operation allows the measurement system to collect, analyze, and publish data at all times. Thus, whenever the measurement system is powered and the software is operational, it collects and analyzes any available data and attempts to derive and display as much useful information as possible about its current condition and the gear set being tested (if any). In the present invention, the always-on and contract functions are not mutually exclusive and can be exercised simultaneously. Data recorded during continuous monitoring can be accessed for subsequent contract testing without requiring additional movement of the gear set.

[0145] An always-on system starts by continuously monitoring the available sensors. It then processes and displays the resulting measurements and data in real time and in a useful format for the user. An always-on system attempts to be as independent and robust as possible in order to intelligently derive a set of parameters, settings, and values ​​that describe its current condition and report discrepancies rather than being disabled or confused by them.

[0146] For example, if a gear set (e.g., a bevel pinion and ring gear) with an 11:38 gear ratio is installed and rolling in a machine, but the software system is set to expect 9:38, the conventional measurement system will not process any information or detect any discrepancies until the contract test has begun. At that time, it will most likely report a fault and / or produce an incorrect measurement result due to not meeting the expected ratio.

[0147] On the other hand, a fault-tolerant always-on system will recognize at the earliest possible moment that both spindles are turning, that they are closely following a given kinematic ratio (and therefore it will infer that the gear set is rolling), and that the kinematic ratio is, for example, 3.455, and that 11:38 is the most likely corresponding actual gear ratio. It will immediately begin analyzing and publishing typical single flank test (SFT) and structure-borne noise (SBN) data associated with such a gear set, including mesh harmonic amplitudes, runout components, and more. By immediately publishing the detected ratio, the machine controller is able to compare the measurement system data with its own expected parameters at an earlier time and expediently initiate a message to the user.

[0148] During always-on operation, follow the Figure 1 The example of the MTE signal depicts a process that continuously monitors the encoder output. For each defined channel, the method observes an encoder stream of frames for that channel. As will be described below, the encoder positions are initially sampled regularly in time, but measurements are often required at regular rotational positions. Therefore, after each frame is received, it can be resampled by local interpolation to maintain regular intervals in the angular position of one of the gear set members. The resulting encoder data is then used to determine an estimate of the current gear ratio by linear regression. If no coordinated motion is detected, the data is discarded and the method waits for the next frame of encoder information. Otherwise, the MTE is calculated from the frame. If the gear ratio is different from expected or appears to have changed from the previous frame, the newly detected ratio is provided to a subsequent analysis module or other interested observer. The MTE is low-pass filtered and downsampled to mitigate quantization noise in the encoder hardware, and the resulting signal can be used for subsequent processing.

[0149] Additional signals may be collected simultaneously with the MTE as a function of the output of other sensors attached to the machine. For each channel, these signals will be grouped into a frame duration that coincides with the frame representing the MTE.

[0150] Each channel adds new frames as they become available; frames collected before the scheduled channel duration are discarded.

[0151] In a preferred implementation of the always-on analysis, a gear set consisting only of a single pinion meshing with a single gear will consist of a meshing channel with a frame duration of a certain number of meshing cycles, a pinion channel with a frame duration of a certain number of pinion revolutions, a gear channel with a frame duration of a certain number of gear revolutions, and a track channel with a frame duration corresponding to one or more complete tracks of the gear and pinion.

[0152] Each channel waits for a new frame, a newly discovered gear ratio, or a change in the configuration parameters of the harmonic decomposition (described below) to perform automatic analysis, such as Figure 2 An example of MTE analysis is depicted in Figure 1. When notification of a gear ratio change is received, all currently retained MTE data is discarded, since appending MTE data from two different gear ratios would not produce useful analysis. When a channel receives new MTE data, that data is appended to the channel record, and potentially "outdated" data (data that falls outside the configured duration of the channel) is discarded. The currently retained MTE frame is passed to the decomposition module, which produces a model of the predictable components in the actual MTE, and the model is published to further analysis or visualization components. The decomposition is recomputed whenever an automated process or user request changes the decomposition settings for a channel. Although not depicted in the figure, additional automated analysis may occur within each channel.

[0153] Figure 3 The basis decomposition process is depicted. The gear ratio, a set of desired harmonic components ("harmonic count"), and optionally the maximum polynomial order are used as inputs to a basis module, which determines a set of basis functions ("linear system") of the signal. The decomposition module contains an algorithm that generates a series of basis coefficients representing the model from the linear system and the input signal as input. The linear system and the corresponding model coefficients are provided as input to an expansion module, which generates an expanded model of the signal ("MTE model"). The difference module subtracts the signal model from the input signal to produce a residual.

[0154] The always-on system is constantly checking for other issues such as tooth damage (such as nicks and burrs), tooth spacing errors, signs of component slippage or creep in the mandrel, intermittent floating or snapping of lightly loaded components, or other abnormal conditions. As data and conclusions about such phenomena are created, it publishes these results to the user interface (UI) and other system components so that appropriate actions can be taken. By doing all this without explicit specific contracts to check for these conditions, monitoring is simplified, phenomena are identified earlier, users are more aware of current conditions regardless of the status of specific contract tests, and the system operates in a more intelligent and autonomous manner.

[0155] The always-on user interface also allows for interaction and exploration. The user is able to navigate through screens and displays, access more detailed views of aspects of measurements, and change real-time parameters and settings for data collection and analysis in order to understand and quantify aspects of the gear set's behavior. To support such extended exploration, the machine side of the tester that controls the parameters of the test (including motion, speed, and load) can be placed in a "hold" state to facilitate closer examination of the data using the always-on analysis and UI capabilities. The hold state can involve pausing or ignoring the flow of new signals and data, while the motion of the gear set continues or stops completely in its previous state. The hold state can also employ a steady stream of new information obtained by scrolling the gear set at a constant speed to allow the operator to view real-time deviations from ideal gear set behavior indefinitely.

[0156] Always-on mode also provides for capturing current data for a predefined duration in a database, file system or volatile computer memory. Users can choose to retain specific intervals of capture and analysis results indefinitely or apply contractual tests retroactively to the retained measurement data.

[0157] In production mode, setup-defined contract tests are automatically initiated by the machine controller according to its predefined cycle sequence and timing, data collected, and results displayed, without interfering with any always-on functions that may be simultaneously active.

[0158] Several approaches can be combined when processing data, such as motion transfer error data derived from rotational SFT measurements. Gear set analysis typically uses "rotational" sampling, in which the gear and pinion encoder signals are uniformly sampled across the rotational variation of the pinion encoder (i.e., the difference between any two consecutive samples of the pinion encoder is constant). The resulting analysis inherently associates periodic phenomena with specific points in the rotation of the gear or pinion being measured. Because the sampling hardware typically uses regular time intervals when sampling, and the gears may rotate at non-constant speeds, the encoder signals must be interpolated to convert regular time samples into regular rotational samples. The bandwidth of the sampling hardware is typically much higher than the bandwidth required to analyze the gear set. Sampling is preferably performed at the highest supported frequency to achieve two main benefits. First, the interpolation responsible for converting time samples into rotational samples has smaller errors. Second, applying a low-pass filter in the data acquisition software and downsampling the oversampled signal reduces the quantization noise present in the digitized signal.

[0159] When the time-sampled encoder signals are converted to rotational samples, the timing information inherent to the time samples is preserved in the form of time stamps associated with each sample. These time stamps provide information about the speed and acceleration of the gear set, which can be used for alternative analysis or conversion back to time-sampled signals. Sampling at the highest possible rate ensures that these time stamps, and the resulting estimated speed and acceleration, have minimal error.

[0160] In contrast to current measurement systems, the present invention maintains a tooth index system that allows the specific temporal characteristics of motion transfer errors to be associated with the contact period between specific gear and pinion teeth. By default, any recording of a gear set test (starting when the gear and pinion are stuck in the rolling tester and begin rolling in a coordinated motion) assigns a gear tooth index of 1 and a pinion tooth index of 1 to the first tooth in contact on the corresponding gear set member. Both the gear and pinion indices increase as the motion continues through each subsequent tooth contact theory cycle, where the index of the member wraps around to 1 whenever the increased index would exceed the number of teeth on each member. The tooth contact theory cycle represents a simplified view of tooth contact, where a single pair of contacting teeth (the teeth currently indexed) is designated as active at any given time. The tooth contact theory cycle can be understood as one rotational period of a gear member divided by its number of teeth. In practice, due to gear design and tooth deformation under load, more than one pair of meshing teeth may be in meaningful contact for a given gear pair at a given time. The tooth indexing system of the system of the present invention preferably assumes this theoretical periodicity, but it is conceivable that it could also support the specification of multiple (overlapping) tooth pairs if sufficient information about the behavior is provided.

[0161] The control interface of the rolling tester can increase the understanding of tooth numbering in two ways. First, the interface can present an "indicator" option that will allow the operator to select the desired index of the gear and pinion teeth and cause the rolling tester to place the desired teeth independently in a predetermined orientation. Second, the interface can present a "teach" option to enable the identification of the desired index sequence. The operator can manually position the gear and pinion to align the specified starting teeth (each starting tooth has an index of 1) in a predetermined orientation, and guide the software through the teaching interface to record the component orientation as a reference. Until the component is later unstuck, all data records provided for analysis or visualization after the teaching operation will include additional labels to indicate the marked reference positions. These reference position labels can be used as a reference, which allows the operator to find which specific teeth are contacted at any point throughout the test.

[0162] Many of the analyses performed can be combined with specialized visualizations that allow the rolling tester data to be explored in novel ways. However, the combination of analysis and visualization is a convenience, not a requirement. In the field of gear set testing, each of the analysis steps and visualizations can be considered separately. Basic analysis can be applied to always-on or contract data without presenting a visualization. The results of these analyses can be compared to thresholds or reduced by some means (e.g., statistically) to facilitate simpler reporting or provide automated make-or-break decisions about the gear set under test.

[0163] At least five types of visualizations are envisioned:

[0164] Model decomposition,

[0165] DFT (Discrete Fourier Transform) amplitude,

[0166] Coverage map,

[0167] Spectrogram, and

[0168] Mesh grid.

[0169] The five types of visualization are preferably all linked by a core model and residuals, which provide the basis for all visualizations.

[0170] In such Figure 4 and 5 In the model decomposition visualization shown, for each selected measurement, the decomposition view shows the expansion of each component, external and non-periodic basis separately on a common axis (where the input gear position represents the independent variable and the measurement represents the dependent variable) as a linear combination of functions in the basis weighted by the determined model coefficients. The residuals for each measurement are shown together with each basis expansion.

[0171] User interface controls allow the operator to toggle visibility of each basis extension or residual.

[0172] The control allows the operator to select a preferred display scale as the fundamental period of any component or external basis represented in one or more models, which causes the input gear position to be scaled so that one unit on the display axis represents the selected fundamental period.

[0173] User interface elements ("tool tips") allow the operator to position the cursor on the graph to discover the specific amplitude of each displayed component and the position of the input gear near the cursor. Figure 8 An example of a tool tip is shown in , where it should be understood that the data shown is only relevant to that particular line and position on that particular graphical display.

[0174] On a separate axis, something like Figure 5The bar chart or line graph depicted in shows the magnitudes of some or all of the model coefficients, which are grouped by measure and basis and labeled to allow specific coefficient amplitudes to be discovered.

[0175] Tool tips allow the operator to position the cursor on the magnitude graph to discover the specific numerical value of a specific coefficient, as well as sufficient information to identify the specific component being queried.

[0176] The decomposition analysis represents the periodic components of the motion transfer error as multiples of harmonics per gear, per pinion, or per mesh, and optionally a non-periodic polynomial trend. Harmonic analysis is common in current gear set test systems. However, the decomposition performed by the method of the present invention improves upon conventional analysis by employing the aforementioned decomposition method, which reduces the effects of spectral leakage and accounts for non-periodic behavior not represented in current systems.

[0177] For each gear in the gear set, the low-order harmonic components determined by the decomposition are displayed to the user as idealized or modeled gear responses, and some amount of these components (selected explicitly by the user or automatically by the program) is subtracted from the motion transmission error to produce a residual representing noise sources unrelated to the rotation of the gear set as well as the higher frequency per-mesh harmonic content. The residual can be the basis for further analysis and visualization.

[0178] In such Figure 6 In the DFT magnitude visualization shown, a bar graph or line graph shows the magnitude of the discrete Fourier transform (DFT) of the residuals of the current model or models as a function of the rotational frequency of the input gear.

[0179] The control allows the operator to select a preferred display scale as the fundamental frequency of any component or external basis represented in one or more models, which results in the rotational frequency of the input gear being expressed in units of the selected frequency.

[0180] The control allows the operator to enable a "virtual orbit" which causes the current data record to be zero-filled to a duration equal to the next larger integer orbit of any two interconnected gears in the gear set so that the rotational frequency calculated in the DFT is consistent with the fundamental rotational frequency of the two members and the meshing frequency.

[0181] When the amplitudes are displayed as a bar graph, the bars for each DFT frequency bin are preferably individually colored, shaded, labeled, or otherwise distinguished according to their correspondence to multiples of the fundamental frequency of each component or external basis (or lack of correspondence with any such basis) to allow easy visual identification of correlations between periodic phenomena and frequencies attributable to the various components of the gear set.

[0182] Tooltips allow the operator to position the cursor on the DFT plot to discover the magnitude and center rotation frequency in a specific DFT frequency bin.

[0183] In such Figure 7 , 8 , 9 and 10, the controls allow the operator to select a preferred display scale as the fundamental period of any component or external basis represented in one or more models, which results in the residuals of the current one or more models being subdivided into "slices" consisting of either:

[0184] · disjoint, contiguous intervals of duration equal to the chosen elementary period, or

[0185] • Duration equal to the overlap interval of the selected elementary cycle plus an additional overlap interval, where consecutive segments overlap by an overlap interval selected automatically by the software or manually by the operator.

[0186] The line graph for each selected measurement shows on a common axis all subdivided residual segments for that measurement (wherein each segment is preferably differentiated by color or style) as a function of the rotational position of the input gear relative to its position at the start of the segment, to allow for examination of variability between data segments that is not attributable to the determined model.

[0187] The tool tip allows the operator to position the cursor over one or more lines to identify the specific interval over which the segment is defined as a function of the rotational position of the input gear.

[0188] The control allows the operator to toggle visibility of the average residual over all displayed segments for each selected measurement. Figure 7 , 8 , 9 and 10, the mean residual is depicted as a black overlay on each segment.

[0189] The controls allow the operator to switch between Figure 8 Visibility of the depicted line or band indicating the point-wise standard deviation or multiples thereof selected by the user control over all displayed segments.

[0190] exist Fig.11 and 12 In the spectral overlay visualization of , the individual segments subdivided as described above for the location overlay view are converted into spectral amplitudes by computing the discrete Fourier transform of each segment. The amplitudes of the DFT for each segment are juxtaposed on a single set of axes, which associate specific DFT frequency bins with corresponding amplitudes for each data segment with the aid of bar or line graphs.

[0191] The selection of the duration of each spectral overlay segment may be made automatically (eg, by appending the duration to the duration selected for the location overlay display), or by specific user selection.

[0192] Tooltips allow the operator to position the cursor over one or more lines to identify the specific interval that defines the segment, along with detailed information on the frequency and amplitude corresponding to the cursor position.

[0193] Controls allow the operator to toggle visibility of the average amplitude computed as the average of the spectral amplitudes or the amplitude of the complex (phase) average of the DFT coefficients for each selected measurement, as well as other statistical measures in each DFT bin (e.g., the first, second, and third quartile points of the spectral amplitude).

[0194] Controls allow the operator to toggle visibility of box and whisker plots for each selected measurement, which present visual information about the statistical variation between DFT amplitudes across bins. The lower and upper bounds of the box depict the first and third quartiles, respectively, of the amplitude distribution in each DFT frequency bin; the center line depicts the median. The whiskers extend above and below the box for multiples of the interquartile range (the difference between the third and first quartiles), configurable by user controls, to depict a range of "typical" amplitudes. Fig.12 As depicted, the individual magnitude indicators may be hidden for those values ​​within the range of the box and whisker plot to provide clarity.

[0195] In an "Overlay Chart" analysis, the residuals are divided into segments of duration equal to a single number of revolutions of one of the components or a contact period between any two teeth. The resulting segments are overlaid on the same axis to allow comparison of relative characteristics between multiple related intervals throughout the test. Summary statistics of the overlapping windows (including mean response, per-sample standard deviation, range of values, and RMS level) provide a characterization of the results that can be used to compare multiple gear sets and to identify anomalous behavior in specific areas of each gear set. Visualization of the Overlay Chart analysis provides a means to identify individual segments and summary statistics.

[0196] exist Fig.13 In the spectrogram visualization of , an image (as a 2D color or grayscale map or a 2D or 3D waterfall plot) for each selected measurement shows the magnitude of the Short Time Fourier Transform (STFT) of the currently determined residual for that measurement.

[0197] Controls allow the operator to select the display scale as the fundamental frequency of any component or external basis represented in the model, which causes the rotational frequency and position of the input gear to be scaled in units of the selected frequency and period, respectively.

[0198] Controls allow the operator to select the duration of the sliding position window of the STFT, measured as either the absolute width or full-width half-maximum as a function of the window.

[0199] The offset between consecutive STFT position windows can be determined by manual selection or by automatic determination so that the window duration, overlap and function meet the constant overlap-add condition required to ensure STFT reversibility within predetermined tolerances.

[0200] As an alternative to manually selecting the duration of the STFT position window, the duration together with the offset may be automatically selected to provide position and spectral resolution of a predetermined aspect ratio.

[0201] Controls allow the operator to enable frequency and position reallocation, which transforms the STFT image into a two-dimensional or three-dimensional scatter plot in which the individual amplitudes of the STFT are shifted to irregular locations derived from the phase information in the STFT.

[0202] The bar graph or line graph shows the periodogram calculated from the currently displayed STFT ( Fig.14 ), as the mean and standard deviation of the frequency response over all position segments in the STFT.

[0203] Controls allow the user to select a spectral level, as the fundamental frequency of any component or external basis represented in the model, such that the frequency is greater than the spectral resolution of the periodogram (which will be used to define contiguous but non-intersecting frequency bands of width equal to the selected frequency level), centered on integer multiples of the frequency of the selected level; at each frequency band, the rms amplitude of the periodogram is calculated and displayed as a bar graph or stepped line on the same axes as the periodogram.

[0204] A short-time spectral analysis is performed on the residuals to highlight the long-term variations of the meshing harmonics. Based on a selected time level that can be selected by the user or by the software based on the test duration and the user's desired spectral bandwidth, a Gabor transform (a short-time Fourier transform with a Gaussian time window) is applied to the residuals and displayed as a two-dimensional spectrogram image with spectral amplitudes varying along columns and time slices varying along rows. The time interval between consecutive columns in the short-time Fourier transform is selected based on the selected Gaussian window to avoid loss and redundancy of temporal information.

[0205] The periodogram (average spectral amplitude calculated over all columns in the spectrogram) shows the typical spectral response of the entire test, as well as the RMS amplitude calculated over a band equal to a single gear or pinion revolution or a single meshing contact cycle. Comparison of individual spectral amplitudes to the RMS level in a wider neighborhood provides a measure of the power concentration near the harmonic of interest. Short-time spectral analysis also provides the basis for more detailed investigations. For example, the time-varying average frequency near each meshing harmonic can be calculated from the spectrogram to quantify phase drift, which will artificially reduce the spectral amplitude reported by traditional testing.

[0206] In such Fig.15In the grid visualization shown, an image of each selected measurement (as a two-dimensional color or grayscale map) shows the statistical variation, where the statistic of interest is selected by user interaction across the mesh duration by mapping the statistic for each mesh cycle to the interaction of a specific tooth index on each of the two interconnected gears in the gear set.

[0207] The residuals of the current model of one or more measurements are subdivided into disjoint, connected segments having a duration equal to a single meshing cycle.

[0208] A reference signal is defined for each residual, either as a point-wise average over all segments, or as an expansion of a basis of only the meshing period components of the model.

[0209] Tooltips allow the operator to position the cursor over an individual meshing period on the mesh to discover the associated segment index, the tooth index of the two interconnected members to which the segment period is mapped, and the numerical value of the selected statistic.

[0210] Separate graphs (each as a two-dimensional graph consisting of a single row or column, or as a bar graph or line graph) respectively show the value of a selected statistic as a function of the tooth index of each interconnected member in the gear set, wherein the statistic is averaged over all teeth of the conjugate members contacted by the teeth of the member associated with the average display.

[0211] Controls allow the operator to select any set of meshing cycles from the mesh grid view for further inspection.

[0212] Detailed view ( Fig.16 ) can be viewed with the grid Figure 1 The grid view allows one or more measurements from a selected meshing cycle to be displayed as a function of the input gear position relative to the start of each segment.

[0213] On the same axis as the measurements of the selected meshing cycle, the detailed view shows the reference signal derived for these measurements.

[0214] In the "grid grid" analysis, the residuals are divided into segments whose duration is always equal to a single contact period between two teeth. Each segment is mapped to a specific gear and pinion tooth that was in contact during its duration. A reference signal is calculated as the average of all per-tooth signals or as an ideal reconstruction of a configurable number of meshing harmonics determined from the decomposition. Summary statistics are calculated for each meshing interval:

[0215] Peak-to-peak amplitude;

[0216] The spacing error is the signal delay that maximizes the cross-correlation between the selected reference signal and the error;

[0217] The RMS (root mean square) error between the residual and the selected reference;

[0218] · Pearson correlation coefficient between the selected reference and the residuals; or

[0219] • One of multiple measures of waveform similarity between a reference and an error.

[0220] The grid visualization allows display of selected summary statistics on a two-dimensional grid indexed by gear tooth number in one dimension and by pinion tooth number in the other. Additional rows and columns provide the average of the selected statistics, respectively, when the gear or pinion tooth index is held constant. One or more intervals in the grid can be selected to populate a detailed view showing the corresponding actual residuals overlaid on the same axis as the currently selected reference waveform.

[0221] Each visualization provides a mechanism to adjust the core model or models that are displayed.Data may be provided by any previously determined analysis channel or by a "contract" channel consisting of measurement records of predetermined length collected previously or in real time.

[0222] User interface controls allow the operator to select any connected subset ("range") of the complete data record on which to compute the model exclusively.

[0223] Controls allow the operator to downsample selected data records to speed processing by applying low-pass filtering and decimation or local or global polynomial interpolation.

[0224] Controls allow the operator to independently select the desired number of functions in each component, external, or aperiodic basis used in the model.

[0225] If measures other than motion transfer error are available in the data record, controls allow the operator to select a preferred measure (or measures) for visualization.

[0226] From the selected range and downsampling rate, a core model and residual signal are derived for each selected measurement using the indicated number of basis functions within each series.

[0227] Operators can group multiple visualizations into linked "sessions," or can treat each visualization as its own independent session.

[0228] The network communication channel enables coordination between multiple views of a single session, whereby operator settings and portions of the signal identified by the operator are used for closer scrutiny. When the operator adjusts settings or highlights portions of the displayed signal in one visualization, the corresponding settings are applied and the selection is highlighted in all other visualizations within the same session.

[0229] Although the present invention has been described with reference to the preferred embodiment, it should be understood that the invention is not limited to the details thereof. The present invention is intended to include modifications obvious to those skilled in the art without departing from the spirit and scope of the appended claims.

Claims

1. A method for testing a gear, the method comprising: providing a gear pair including a first gear and a second gear operable to roll in meshing engagement with each other, rolling the first gear and the second gear together on a gear testing machine, the gear testing machine being configured with one or more sensors operable to provide output signals indicative of rolling characteristics of the one or more gears, generating at least one independent signal based on at least one of a signal output from the one or more sensors or a signal derived from analysis, At least one frame portion having a non-zero duration is selected from the at least one independent signal, wherein each of the selected at least one frame portion is processed according to the following steps: determining a set of model basis functions based on the selected frame portion, projecting the selected frame portion onto the model basis functions to generate model coefficients, determining a reconstruction model from the model coefficients and the model basis functions, subtracting the reconstructed model from the selected frame portion to produce a residual, The method further comprises at least one of the following: (a) displaying and analyzing the residual to identify quality characteristics of at least one of the first gear and the second gear, as well as (b) displaying and analyzing the model coefficients or reconstructing the model to identify quality characteristics of at least one of the first gear and the second gear, in, The method further comprises: accepting or rejecting the at least one of the first gear and the second gear based on the quality characteristic, Therein, a plurality of said independent signals are sampled simultaneously at consistent intervals to facilitate direct correlation and comparison.

2. The method according to claim 1, characterized in that The model basis functions are determined from at least one of the following: (a) the duration of the selected frame portion, (b) a gear ratio between the first gear and the second gear, (c) a fundamental harmonic frequency generated by the rotation of the first gear and the second gear or a duration of a meshing period in which the first gear and the second gear are meshed, (d) a set of predetermined harmonics of each fundamental frequency, said set including zero, and (e) a set of non-periodic basis functions defined over said duration of said selected frame portion, said set comprising zero.

3. The method according to claim 1, characterized in that The reconstructed model and the residuals are continuously re-evaluated as additional information is obtained from the sensors during machine operation.

4. The method according to claim 1, characterized in that: The at least one independent signal is collected at uniform time samples and subsequently interpolated to produce a sequence of samples regularly spaced in the position of one of the first gear and the second gear.

5. The method according to claim 1, characterized in that: Display residuals, rebuild model, or a combination of these: (a) subdividing the selected frame portion into a plurality of contiguous signal segments of equal duration, wherein the duration is equal to one or more of:

1. One or more complete meshing cycles, 2. One or more complete revolutions of any gear in the gear set, (b) representing the signal segment as a function of one or more of:

1. Time, 2. The position of one of the gears in the gear set, 3. Frequency, with the help of Fourier transform, (c) presenting the signal segment functions aligned and overlaid on the same independent axis as one or more of the following:

1. The trace of the signal fragment function, 2. The trace of the statistical variation of the signal segment function, consisting of the arithmetic mean and / or variance, determined point by point.

6. The method of claim 1, wherein displaying the residual, reconstructing the model, or a combination thereof comprises: (a) subdividing the selected frame portion into a plurality of contiguous signal segments having equal duration, wherein the duration is equal to one or more complete meshing cycles, (b) determining for each signal segment a scalar metric for each signal segment consisting of one or more of the following:

1. Peak-to-peak variation, 2. RMS amplitude, 3. a temporal or rotational shift of said signal segment required for optimal alignment of said signal segment with a reference, determined as either: an average of at least two segments or an average of a subset of said reconstructed models having the same duration as said signal segment, 4. A similarity measure to a reference, determined as any of: said average of at least two segments or said average of a subset of said reconstructed models having said same duration as said signal segments, said measure consisting of one or more of: Pearson correlation coefficient and / or root mean square error, (c) Each fragment is assigned two indices, the first index corresponding to any of the following:

1. the number of complete revolutions of one of the first gear or the second gear observed in the segment signal before the segment duration begins, or 2. The number of meshing cycles observed before the segment duration begins, modulo the number of teeth on the one of the first gear or the second gear; the second index corresponds to either:

3. the number of complete revolutions of the other of the first gear or the second gear observed in the segment signal before the segment duration begins, or 4. said number of engagement cycles observed before said segment duration begins, modulo the number of teeth on the other of said first gear or said second gear, (d) presenting the scalar metric mapped to a range of gray levels or color values ​​onto a two-dimensional image, wherein one coordinate of the image corresponds to the first index and another coordinate of the image corresponds to the second index.

7. The method according to claim 1, characterized in that Display residuals, rebuild model, or a combination of these: (a) subdividing the selected frame portion into a plurality of contiguous or overlapping signal segments of predetermined equal duration, each signal segment being sequentially indexed, (b) representing the signal segment as a spectral function by means of Fourier transformation, (c) determining from the spectral function one or more of the following measures:

1. Amplitude, 2. Logarithmic amplitude, 3. Phase, 4. Phase drift, which is the difference between the actual phase and the phase predicted by advancing the phase of the previous spectral function by the time or rotation interval between the two segments, (d) presenting the spectral measurement having values ​​mapped to a series of gray levels or color values ​​as a two-dimensional image, wherein one coordinate of the image corresponds to the index of the segment and another coordinate of the image corresponds to the frequency variable of the spectral function.

8. The method according to claim 1, characterized in that Display contains: A trace of one or more of the residual, the reconstructed model, or a combination thereof on a graph wherein an independent axis represents time or position of one of the first gear or the second gear in a gear set and a dependent axis represents amplitude of the displayed quantity.

9. The method according to claim 1, characterized in that: Display model coefficients including: Plot one or more of the following: (a) a bar graph with one bar per coefficient and a height proportional to the magnitude of the coefficient; (b) one or more radial graphs on which are drawn line segments, one for each coefficient, radiating from the coordinate origin and having a length proportional to the complex amplitude of the corresponding coefficient and an angle with the horizontal axis proportional to the complex phase of the corresponding coefficient.

10. The method according to claim 1, characterized in that The display and analysis results are retained on an electronic storage medium for recall, wherein the display and comparison of more than one previous record is accomplished by one or more of the following: (a) Simultaneous adjacent presentation, (b) sequential presentation with a progression guided by user interaction, (c) Sequential presentation with automatic, animated progression.

11. A method for testing a gear, the method comprising: providing a gear pair including a first gear and a second gear operable to roll in meshing engagement with each other, rolling the first gear and the second gear together on a gear testing machine, the gear testing machine being configured with one or more sensors operable to provide output signals indicative of rolling characteristics of the one or more gears, generating at least one independent signal based on at least one of a signal output from the one or more sensors or a signal derived from analysis, At least one frame portion having a non-zero duration is selected from the at least one independent signal, wherein each of the selected at least one frame portion is processed according to the following steps: determining a set of model basis functions based on the selected frame portion, projecting the selected frame portion onto the model basis functions to generate model coefficients, determining a reconstruction model from the model coefficients and the model basis functions, subtracting the reconstructed model from the selected frame portion to produce a residual, The method further comprises at least one of the following: displaying and analyzing the residual to identify a quality characteristic of at least one of the first gear and the second gear, as well as displaying and analyzing the model coefficients or reconstructing the model to identify mass characteristics of at least one of the first gear and the second gear, in, The method further comprises: accepting or rejecting the at least one of the first gear and the second gear based on the quality characteristic, The method proceeds in an always-on mode, wherein the previous steps are repeated and continuously performed as long as the one or more sensors detect coordinated motion between the gear pairs and a new data frame can be populated.

12. The method according to claim 11, characterized in that In the always-on mode, the gear ratio of the first gear and the second gear being scrolled is not known in advance, but the always-on mode determines the gear ratio based on the real-time data it receives.

13. The method according to claim 11, characterized in that A user is able to view information including at least one of measurement, display, visualization, and results at any time while the always-on mode is operating, the information being available regardless of whether any other test mode is simultaneously active.

14. The method according to claim 13, characterized in that The information comprising at least one of a measurement, a display, a visualization, and a result comprises at least one of: (a) the amplitude and phase of one or more harmonics of a frequency associated with the meshing period; (b) Average signal shape over the duration of one meshing cycle; (c) gear runout harmonic amplitude; (d) Average signal shape of gear runout; (e) the presence and location of tooth lesions; (f) removing the average signal shape of the pinion base; and (g) Average signal shape after removing the gear base.

15. The method according to claim 11, characterized in that It also includes interactively manipulating the display to better examine various aspects of the displayed data.

16. The method according to claim 11, characterized in that The display can be paused so that the current set of displayed data can be examined without being replaced by a subsequent frame.

17. The method according to claim 11, characterized in that A current set of data can be captured and stored for later retrieval or verification and / or for comparison with current data.

18. The method according to claim 11, characterized in that Differences between current data and stored reference or master data can be checked and visualized.

19. The method according to claim 11, characterized in that The always-on mode also includes identifying and locating tooth damage on at least one of the first gear and the second gear without the need to separately establish any test parameters or settings, the always-on mode identifying the component rotation or tooth index that best represents the location of the identified damage, and the always-on mode automatically notifying the user of such damage via a text message and / or graphical display.