Device health estimation by combining contextual information with sensor data

Abstract

A method and system for detecting fault in a machine. During operation, the system obtains control signals and corresponding sensor data that indicates a condition of the machine. The system determines consistent time intervals for each of the control signals. During a consistent time interval the standard deviation of a respective control signal is less than a respective predetermined threshold. The system aggregates the consistent time intervals to determine aggregate consistent intervals. The system then maps the aggregate consistent intervals to the sensor data to determine time interval segments for the sensor data. The system may generate features based on the sensor data. Each respective feature is generated from a time interval segment of the sensor data. The system trains a classifier using the features, and applies the classifier to additional sensor data indicating a condition of the machine over a period of time to detect a machine fault.

Description

FIELD

The present disclosure generally relates to estimating device health. More specifically, the present disclosure relates to a method and system for accurately diagnosing device failure by combining a device's operating context with sensor data.

RELATED ART

The growing Internet of Things is predicted to connect 30 billion devices by 2020. This will bring in tremendous amounts of data and drive the innovations needed to realize the vision of Industry 4.0, which includes cyber-physical systems monitoring physical processes, and communicating and cooperating with each other and with humans in real time. One of the key challenges is how to analyze large amounts of data to provide useful and actionable information for business intelligence and decision making. In particular, one challenge is to prevent unexpected downtime and its significant impact on overall equipment effectiveness (OEE) and total cost of ownership (TCO) in many industries. Continuous monitoring of equipment and early detection of incipient faults can support optimal maintenance strategies, prevent downtime, increase productivity, and reduce costs. To that end, there have been a number of anomaly detection and diagnosis methods proposed for detecting machine fault and estimating machine health.

Some have proposed applying different approaches to detect anomalies for various types of equipment, including statistical methods, neural network methods, and reliability methods. Some approaches focus on analyzing, combining, and modeling sensor data (e.g. vibration, current, acoustics signal) to detect machine faults. However, in some cases, these approaches may generate false alarms. Previous approaches have also used vibration data and/or acceleration data for diagnosing machine imbalance fault conditions. Other approaches may use temperature data to diagnose faults. However, it can be expensive to acquire vibration data and it may be difficult and insufficiently accurate to use temperature data to perform diagnostics.

SUMMARY

One embodiment of the present invention provides a method for detecting fault in a machine. During operation, the system obtains a plurality of control signals associated with controlling the machine and sensor data that indicates a condition of the machine during a time period when the plurality of control signals control the machine. The system then determines consistent time intervals for each of the plurality of control signals, in which during a consistent time interval the standard deviation of a respective control signal is less than a respective predetermined threshold. The system may aggregate the consistent time intervals of the plurality of control signals to determine aggregate consistent intervals. The system then maps the aggregate consistent intervals of the plurality of control signals to the sensor data to determine a plurality of time interval segments for the sensor data. The system may generate a plurality of features based on the sensor data, in which each respective feature is generated from a time interval segment of the plurality of time interval segments for the sensor data. The system may then train a classifier using the plurality of features, and subsequently apply the classifier to additional sensor data indicating a condition of the machine over a period of time to detect a machine fault.

In one variation on this embodiment, the plurality of control signals includes spindle motor speed, spindle load, and actual spindle speed, and the sensor data is temperature data indicating a temperature associated with the machine.

In one variation on this embodiment, aggregating the consistent time intervals includes determining an intersection of sets of consistent time intervals over all control signals.

In one variation on this embodiment, generating the features includes computing an average, a standard deviation, a maximum fast Fourier transform (FFT) value, and a FFT frequency at maximum amplitude for the sensor data.

In a further variation on this embodiment, the generated features form a high-dimensional feature space. The system also applies principal component analysis (PCA) to project the high-dimensional feature space into a low-dimensional space, and applies linear discriminant analysis (LDA) to determine an optimal coordinate transformation that provides maximum separation between classes.

In a further variation on this embodiment, determining consistent time intervals further includes generating a temporal segment representation of the machine's operation context.

In a further variation on this embodiment, applying the classifier further includes generating features for the classifier with same conditions as in classifier training by determining time intervals of a primary control signal that have same values for the primary control signal as a value of the primary control signal when generating training features.

In a further variation on this embodiment, the system removes one or more control signal intervals that are inconsistent from the plurality of control signals.

BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 presents a diagram illustrating an exemplary machine fault detection system, in accordance with an embodiment of the present invention.

FIG. 2 presents a flow chart illustrating an exemplary process to generate a classifier for detecting machine faults, in accordance with an embodiment of the present invention.

FIG. 3 illustrates an exemplary graph of the result of a temporal segmentation scheme with spindle speed control.

FIG. 4 illustrates an exemplary graph of spindle speed control with consistent time segments.

FIG. 5 illustrates an exemplary diagram of aggregating control signals.

FIG. 6 illustrates an exemplary graph of principal component analysis-linear discriminant analysis results for group 1.

FIG. 7 illustrates an exemplary graph of principal component analysis-linear discriminant analysis results for group 2.

FIG. 8 illustrates an exemplary fault detection apparatus, in accordance with an embodiment.

FIG. 9 presents an exemplary server in a machine fault detection system, in accordance with an embodiment of the present invention.

In the figures, like reference numerals refer to the same figure elements.

DETAILED DESCRIPTION

The following description is presented to enable any person skilled in the art to make and use the invention, and is provided in the context of a particular application and its requirements. Various modifications to the disclosed embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments and applications without departing from the spirit and scope of the present invention. Thus, the present invention is not limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed herein.

The data structures and code described in this detailed description are typically stored on a computer-readable storage medium, which may be any device or medium that can store code and/or data for use by a computer system. The computer-readable storage medium includes, but is not limited to, volatile memory, non-volatile memory, magnetic and optical storage devices such as disk drives, magnetic tape, CDs (compact discs), DVDs (digital versatile discs or digital video discs), or other media capable of storing computer-readable media now known or later developed.

Overview

Embodiments of the present invention solve the problem of diagnosing machine fault in the absence of vibration data by training a classifier to intelligently combine device operating context information and sensor data to detect abnormal conditions and accurately diagnose device failure. This disclosure describes a data fusion method that combines sensor data indicating the condition of a machine tool with the machine's operating context to detect machine faults. A fault detection system may train a classifier to detect machine faults using features generated from the sensor data. One or more sensors can monitor the machine tool and provide streams of sensor data indicating the condition of the machine tool, including sensor data such as temperature data. Contextual data may include data indicating multiple control signals used to control and operate the machine tool.

The system can leverage contextual information to classify sensor data more accurately. By accounting for environmental or external input (e.g., fluctuations in control signals) the system can reduce or eliminate the number of false alarms. For example, the system may diagnose imbalance of a machine tool that has a cutting tool attached to a spindle. The spindle rotates to turn the cutting tool. A machine tool controller may use closed-loop control with multiple control signals to control the machine. Some of these control signals include control signals for spindle motor speed, spindle load, and actual spindle speed. The control signals may fluctuate and that can affect the output or condition of the machine. The system can train the classifier using only temperature data that correspond to those portions of the control signals that are stable and avoid the fluctuating input signals. This improves the accuracy of the classifier when the system applies the classifier to temperature data that corresponds to stable segments of the control signals with the same control signal conditions.

The system determines the time intervals at which the control signals are consistent (or constant) and uses the consistent intervals to segmentize corresponding sensor data. For example, the control signals may be consistent during some interval with a standard deviation equal to or below a predetermined threshold (e.g., a standard deviation of zero). The system defines a temporal segment for each of the consistent portions of the control signals and may generate aggregated segments that each represents an aggregation of the temporal segments. The system then maps the aggregated segments to the temperature sensor data to define segments in the temperature sensor data. The system uses the segments in the temperature sensor data as features to train a classifier. For example, some of the features may include average, standard deviation, maximum fast Fourier transform (FFT) value, and FFT frequency at maximum amplitude for a segment. The system thus extracts features from the temperature sensor data using the contextual information derived from the control signals. The system trains a classifier using the extracted features. The system can use the trained classifier to detect machine faults for sensor data that corresponds to consistent control signal conditions. The classifier can detect (or predict) abnormal or anomalous temperature sensor data coming from the machine, and can alert a machine operator that there may be an current (or upcoming) problem with the machine.

Thus, by considering the machine's operating context, the system can analyze the device's health condition and diagnose device failure using cheaper sensor data, without needing to use vibration data that may be ideal but is more to expensive to acquire.

Exemplary Machine Fault Detection System

FIG. 1 presents a diagram illustrating an exemplary machine fault detection system 100, in accordance with an embodiment of the present invention. Machine fault detection system 100 includes a fault detection server 102 communicating with one or more sensors 104 over a network 106 that can detect the current condition of a machine 108 and provide continuous sensor data that indicates the condition of machine 108. For example, sensor 104 may provide continuous temperature data for machine 108 by measuring and/or detecting the temperature of the machine. Server 102 may also communicate with an agent 110. Agent 110 may communicate with an adapter 112 to retrieve control signal data.

System 100 may use MTConnect to diagnose machine health condition by combining sensor data with operating context information. MTConnect is an open-source communication protocol designed to allow machine tools and other equipment to talk to one another and to computer programs that process data from the machines. MTConnect was developed to connect various legacy machines independent of the controller providers. MTConnect allows for monitoring machine operating context in real-time.

In one embodiment, agent 110 may format control signal data received from adapter 112 into an MTConnect standard XML stream, and respond to HTTP requests by returning the appropriate control signal data. In some embodiments, agent 110 can also send control signal data to server 102 as agent 110 collects the data. Adapter 112 collects and filters data that includes control signal data from a machine tool controller 114, and sends the collected data to agent 110.

Machine tool controller 114 may control machine 108 with closed-loop control and multiple control signals. Examples of control signals include spindle motor speed, spindle load, actual spindle speed and Y drive load. Machine controller 114 may use feedback from the current state of the machine to control machine 108.

Fault detection server 102 may use the sensor data to detect problems with machine 108. Fault detection server 102 may include a sensor data receiver 116, a control signal data receiver 118, a feature generator 120, a classifier generator 122, and a fault detector 124.

Sensor data receiver 116 may receive continuously streaming (e.g., time series) sensor data from sensor 104. Control signal data receiver 118 may receive control signal data from agent 110. Feature generator 120 may analyze a control signal to determine consistent intervals and aggregate time intervals, and map the aggregate time intervals to time series sensor data (e.g., indicating temperature of machine 108) to segmentize the sensor data and generate features. Classifier generator 122 may generate a classifier by training the classifier on the generated features.

Fault detector 124 may apply the generated classifier to sensor data and control signal input (e.g., in the form of temporal segments) in order to detect machine faults. In some embodiments, fault detector 124 may utilize feature generator 120 to generate features from sensor data using control signal input so that the classifier can classify sensor data segments under the same control signal conditions that the classifier is trained with. Further, some implementations may utilize different sensors and/or additional sensors to detect or measure other machine conditions.

System 100 may diagnose any fault with machine 108. For example, system 100 may detect imbalance problems with machine 108. System 100 may determine a degree of imbalance for machine 108. Machine 108 may be a machine tool with a spindle holding a cutting tool. For example, there may be an imbalance problem if the machine tool, spindle, and/or cutting tool is incorrectly positioned.

System 100 can diagnose any fault using a combination of contextual data (e.g., control signal) received from agent 110 and sensor data that complement each other. System 100 can use the contextual data to discretize (e.g., segmentize) the sensor data and train a classifier using features generated from the sensor data segments. System 100 can apply the classifier to detect faults in machine 108, and alert an operator 126.

Generating a Classifier

FIG. 2 presents a flow chart illustrating an exemplary process to generate a classifier for detecting machine faults, in accordance with an embodiment of the present invention. Embodiments of the present invention are not limited to the operations depicted in FIG. 2, and some embodiments may perform the operations of FIG. 2 in a different order or with operations that vary from that depicted in FIG. 2

As depicted in FIG. 2, the system may initially obtain sensor data and control signal data (operation 202). For example, the system may obtain multiple control signals such as a control signal for controlling a spindle motor speed, a control signal for spindle load, and a control signal for actual spindle speed. The system may also obtain continuously streaming (e.g., time series) sensor data such as the temperature data for a machine. In some embodiments, the system can also obtain other time series sensor data from additional sensors or different sensor types not depicted in FIG. 1.

The system may then perform a segmentation technique that provides a temporal representation of the machine's operation context, and combine the temporal representation with sensor data to estimate machine health.

The system may analyze the control signal data to determine consistent time intervals for control signals (operation 204). These are time intervals that have a standard deviation at or below a minimum predetermined threshold (e.g., the predetermined threshold may be zero).

In some embodiments, to utilize the control signals to provide temporal segmentation, e.g., assuming quasi-steady state, the system determines the time intervals in which the following conditions are satisfied: (i) all experiments display same values (e.g., result in same values) for the primary control signal (e.g., actual spindle speed) and (ii) all the control signals are constant over the same period. Note that, to determine the dynamic response, rather than quasi steady state response, the control signals should be consistent across the experiments so that responses are compared under the same set of control inputs. FIG. 3 shows a graph 300 illustrating the result of this temporal segmentation scheme with the spindle speed control. In some embodiments, the classifier classifies sensor data corresponding to control signal input (e.g., consistent time intervals) that satisfies these conditions. For example, the value of the primary control signal is the same for features when applying a classifier and for features when training the classifier. As another example, all control signals are constant over the same time period for features when applying the classifier and for features when training the classifier.

In some embodiments, there may be multiple control signals. For each sample time, the system may compute the standard deviation for each control signal separately (e.g., zero standard deviation shows a control signal which has the same value for all the sample data). In other words, for each of the control signals, the system may compute the standard deviation at each time step (e.g., time interval). The system may identify the periods with standard deviation at or below a predetermined threshold to find the consistent time intervals. For example, the predetermined threshold may be zero. For example, FIG. 4 illustrates a graph 400 showing consistent time intervals as 16 segments along the time axis. The system may remove one or more control signals that are inconsistent (e.g., this assumes that these control signals are secondary feedback signals). In some embodiments, the system may remove one or more inconsistent control signal intervals.

The system can determine the intersection of the sets of consistent time intervals over all (or a plurality of) the control signals to determine the aggregate time intervals (e.g., temporal segments) over which the control signals are statistically consistent (operation 206). This is illustrated by a graph 500 in FIG. 5. For example, the system may determine that there are 16 temporal segments for the control signals, as depicted in the graphs 400, 500 in FIG. 4-FIG. 5.

System 100 may then generate features from segments of the sensor data that correspond to the aggregate time intervals for the control signals (operation 208). The system may map the aggregate time intervals (e.g., temporal segments) that are consistent to the sensor data to segmentize the sensor data. The system may analyze each segment of the time series for sensor data separately, and decompose the sensor data into features using time-domain and frequency domain analysis. For example, for each segment, system 100 may generate features such as the average, standard deviation, maximum FFT value, and FFT frequency at maximum amplitude. These features characterize or summarize observations of the sensor data for a segment. The generated features may form a high-dimensional feature space. For the examples depicted in FIG. 3-FIG. 7 the system may determine that there are 16 temporal segments, and the system may generate a 64 dimensional feature space to diagnose machine imbalance.

In some embodiments, the system projects the high-dimensional data to a much smaller sub-space to prevent over-fitting. The system may use linear transformation-based approaches. The system may use principal component analysis (PCA) to project a high-dimensional feature space into a low-dimensional space followed by a linear discriminant analysis (LDA) to search the optimal separation among various device health conditions. For example, the system may use PCA to reduce the dimensionality from 64 to 4. The system may use LDA to determine the optimal coordinate transformation that provides maximum separation between classes.

FIG. 6 and FIG. 7 illustrates graphs 600, 700 showing the results of PCA-LDA analysis for fluid temperature sensor data. The inventors determined from experimental results that a temperature sensor located at a motor also exhibited diagnostic capability after applying control-based temporal segmentation. Note that besides temperature sensor data, system 100 can use any other suitable streams of sensor data indicating the condition of the machine, such as pressure sensor data. System 100 can generate features from segments using different types or combinations of sensor data. In some embodiments, system 100 can also generate a classifier using other types of control signals and combinations of control signals besides the control signals described in this specification.

The system may then train a classifier based on the features (operation 210). System 100 may use any type of classifier, including examples such as decision tree classifiers, naive Bayes classifiers, support vector machines, rule-based classifiers, and neural networks. System 100 may apply the trained classifier on streaming temperature sensor data indicating the condition of the machine to predict upcoming machine faults or detect current machine faults (operation 212). In one embodiment, the classifier may classify sensor data that corresponds to control signals matching the same conditions as the control signals used in training the classifier.

Exemplary Apparatus

FIG. 8 presents a block diagram illustrating an exemplary apparatus 800 for detecting machine faults, in accordance with an embodiment. Apparatus 800 can comprise a plurality of modules which may communicate with one another via a wired or wireless communication channel. Apparatus 800 may be realized using one or more integrated circuits, and may include fewer or more modules than those shown in FIG. 8. Further, apparatus 800 may be integrated in a computer system, or realized as a separate device which is capable of communicating with other computer systems and/or devices.

Specifically, apparatus 800 can comprise a sensor data receiver 802, a control signal data receiver 804, a feature generator 806, a classifier generator 808, and a fault detector 810. Note that apparatus 800 may also include additional modules and data not depicted in FIG. 8, and different implementations may arrange functionality according to a different set of modules. Embodiments of the present invention are not limited to any particular arrangement of modules.

Sensor data receiver 802 may receive a continuous stream of sensor data from a sensor. The sensor data indicates a condition of a machine, such as a measure of the temperature or pressure of a machine. Control data signal data receiver 804 may receive control signals from an agent, and the agent may obtain the control signals through an adapter connected to the machine. Feature generator 806 may analyze the control signal data to determine consistent intervals and aggregate time intervals, and segmentize corresponding sensor data to generate features. Classifier generator 808 may generate a classifier based on the generated features. Fault detector 810 may apply the generated classifier to sensor data to detect machine faults. Fault detector 810 may utilize feature generator 806 to generate features from sensor data and control signal input so that the classifier can classify sensor data segments under the same control signal conditions and/or other conditions that the classifier is trained with.

FIG. 9 presents a machine fault detection server 900 in a machine fault detection system, in accordance with an embodiment of the present invention. In FIG. 9, server 900 includes a processor 902, a memory 904, and a storage device 906. Storage device 906 stores programs to be executed by processor 902. Specifically, storage device 906 stores a sensor data receiver 908, a control signal data receiver 910, a feature generator 912, a classifier generator 914, and a fault detector 916, as well as other applications, such as applications 918 and 920. Machine fault detection server 900 may be coupled to an optional display 922, a keyboard 924, and a pointing device 926.

Sensor data receiver 908 may receive a continuous stream of sensor data from a sensor. The sensor data indicates a condition of a machine, such as a measure of the temperature or pressure of a machine. Control signal data receiver 910 may receive control signals from an agent, and the agent obtains the control signals through an adapter connected to the machine. Feature generator 912 may analyze the control signal data to determine consistent intervals and aggregate time intervals, and segmentize corresponding sensor data to generate features. Classifier generator 914 may generate a classifier based on the generated features. Fault detector 916 may apply the generated classifier to sensor data and control signal input (e.g., in the form of temporal segments) in order to detect machine faults. In some embodiments, fault detector 916 may utilize feature generator 912 to generate features from sensor data and control signal input so that the classifier can classify sensor data segments under the same control signal conditions and/or other conditions that the classifier is trained with.

The methods and processes described in the detailed description section can be embodied as code and/or data, which can be stored in a computer-readable storage medium as described above. When a computer system reads and executes the code and/or data stored on the computer-readable storage medium, the computer system performs the methods and processes embodied as data structures and code and stored within the computer-readable storage medium.

Furthermore, the methods and processes described below can be included in hardware modules. For example, the hardware modules can include, but are not limited to, application-specific integrated circuit (ASIC) chips, field-programmable gate arrays (FPGAs), and other programmable-logic devices now known or later developed. When the hardware modules are activated, the hardware modules perform the methods and processes included within the hardware modules.

The foregoing descriptions of embodiments of the present invention have been presented for purposes of illustration and description only. They are not intended to be exhaustive or to limit the present invention to the forms disclosed. Accordingly, many modifications and variations will be apparent to practitioners skilled in the art. Additionally, the above disclosure is not intended to limit the present invention. The scope of the present invention is defined by the appended claims.

Claims

1. A computer-executable method for detecting fault in a machine, comprising: obtaining a control signal associated with controlling the machine and sensor data that indicates a condition of the machine during a time period when the control signal controls the machine;determining consistent time intervals for the control signal, wherein during a consistent time interval the standard a standard deviation of the control signal is less than a predetermined threshold;mapping the consistent time intervals to the sensor data to determine a plurality of time interval segments for the sensor data;generating a plurality of training features based on the sensor data, wherein each respective feature is generated in association with from a time interval segment;providing the plurality of training features as input to a classifier to train the classifier to classify abnormal sensor data during a respective consistent time interval;generating new features for the classifier with same conditions as in classifier training by determining time intervals of a primary control signal that have same values for the primary control signal as a value of the primary control signal when generating the training features; anddetecting a machine fault by providing new features associated with additional machine sensor data as input to the classifier to detect abnormal sensor data during a respective consistent time interval.

2. The method of claim 1, wherein the control signal is at least one of a spindle motor speed, spindle load, and actual spindle speed; and wherein the sensor data is temperature data indicating a temperature associated with the machine.

3. The method of claim 1, wherein generating the plurality of training features includes computing at least one of an average, a standard deviation, a maximum fast Fourier transform (FFT) value, and a FFT frequency at maximum amplitude for the sensor data.

4. The method of claim 3, wherein the generated training features form a high-dimensional feature space, further comprising: applying principal component analysis (PCA) to project the high-dimensional feature space into a low-dimensional space; andapplying linear discriminant analysis (LDA) to determine an optimal coordinate transformation that provides maximum separation between classes.

5. The method of claim 1, wherein determining consistent time intervals further comprises generating a temporal segment representation of the machine's operation context.

6. The method of claim 1, further comprising: removing one or more control signal intervals that are inconsistent from a plurality of control signals before determining aggregate consistent intervals based on the plurality of control signals.

7. A non-transitory computer-readable storage medium storing instructions which when executed by a computer cause the computer to perform a method for detecting fault in a machine, the method comprising: obtaining a control signal associated with controlling the machine and sensor data that indicates a condition of the machine during a time period when the control signal controls the machine;determining consistent time intervals for the control signal, wherein during a consistent time interval a standard deviation of the control signal is less than a predetermined threshold;mapping the consistent time intervals to the sensor data to determine a plurality of time interval segments for the sensor data;generating a plurality of training features based on the sensor data, wherein each respective feature is generated in association with a time interval segment;providing the plurality of training features as input to a classifier to train the classifier to classify abnormal sensor data during a respective consistent time interval;generating new features for the classifier with same conditions as in classifier training by determining time intervals of a primary control signal that have same values for the primary control signal as a value of the primary control signal when generating the training features; anddetecting a machine fault by providing new features associated with additional machine sensor data as input to the classifier to detect abnormal sensor data during a respective consistent time interval.

8. The storage medium of claim 7, wherein the control signal is at least one of a spindle motor speed, spindle load, and actual spindle speed; and the and wherein the sensor data is temperature data indicating a temperature associated with the machine.

9. The storage medium of claim 7, wherein generating the plurality of training features includes computing at least one of an average, a standard deviation, a maximum fast Fourier transform (FFT) value, and a FFT frequency at maximum amplitude for the sensor data.

10. The storage medium of claim 7, wherein determining consistent time intervals further comprises generating a temporal segment representation of the machine's operation context.

11. The storage medium of claim 7, wherein the method further comprises: removing one or more control signal intervals that are inconsistent from a plurality of control signals before determining aggregate consistent intervals based on the plurality of control signals.

12. A computing system comprising: one or more processors;a memory; anda non-transitory computer-readable medium coupled to the one or more processors storing instructions stored that, when executed by the one or more processors, cause the computing system to perform a method comprising:obtaining a control signal associated with controlling the machine and sensor data that indicates a condition of the machine during a time period when the control signal controls the machine;determining consistent time intervals for the control signal, wherein during a consistent time interval a standard deviation of the control signal is less than a predetermined threshold;mapping the consistent time intervals to the sensor data to determine a plurality of time interval segments for the sensor data;generating a plurality of training features based on the sensor data, wherein each respective feature is generated in association with from a time interval segment;providing the plurality of training features as input to a classifier to train the classifier to classify abnormal sensor data during a respective consistent time interval;generating new features for the classifier with same conditions as in classifier training by determining time intervals of a primary control signal that have same values for the primary control signal as a value of the primary control signal when generating the training features; anddetecting a machine fault by providing new features associated with additional machine sensor data as input to the classifier to detect abnormal sensor data during a respective consistent time interval.

13. The computing system of claim 12, wherein the control signal is at least one of a spindle motor speed, spindle load, and actual spindle speed; and wherein the sensor data is temperature data indicating a temperature associated with the machine.

14. The computing system of claim 12, wherein generating the plurality of training features includes computing at least one of an average, a standard deviation, a maximum fast Fourier transform (FFT) value, and a FFT frequency at maximum amplitude for the sensor data.

15. The method of claim 1, further comprising: aggregating consistent time intervals of a plurality of control signals to determine aggregate consistent intervals.

16. The method of claim 15, wherein aggregating the consistent time intervals comprises determining an intersection of sets of consistent time intervals over all control signals.

17. The storage medium of claim 7, wherein the method further comprises: aggregating consistent time intervals of a plurality of control signals to determine aggregate consistent intervals.

18. The storage medium of claim 17, wherein aggregating the consistent time intervals comprises determining an intersection of sets of consistent time intervals over all control signals.

19. The computing system of claim 12, wherein the method further comprises: aggregating consistent time intervals of a plurality of control signals to determine aggregate consistent intervals.

20. The computing system of claim 19, wherein aggregating the consistent time intervals comprises determining an intersection of sets of consistent time intervals over all control signals.