U.S. patent application number 14/790084 was filed with the patent office on 2016-10-27 for fuel gauge visualization of iot based predictive maintenance system using multi-classification based machine learning.
The applicant listed for this patent is PROPHECY SENSORS, LLC. Invention is credited to Biplab Pal, Amit Purohit.
Application Number | 20160313216 14/790084 |
Document ID | / |
Family ID | 57147600 |
Filed Date | 2016-10-27 |
United States Patent
Application |
20160313216 |
Kind Code |
A1 |
Pal; Biplab ; et
al. |
October 27, 2016 |
FUEL GAUGE VISUALIZATION OF IOT BASED PREDICTIVE MAINTENANCE SYSTEM
USING MULTI-CLASSIFICATION BASED MACHINE LEARNING
Abstract
A method and system of a predictive maintenance IoT system
comprises receiving a plurality of sensor data over a
communications network and determining one or more clusters from
the sensor data based on a pre-determined rule set. Further, the
sensor data is classified through a machine learning engine and the
sensor data is further base-lined through a combination of database
architecture, data training architecture, and a base-lining
algorithm. Intensity or degree of fault state is mapped to a fuel
gauge to be depicted on a user interface and a predictive
maintenance state is predicted through a regression model and
appropriate alarm is raised for user action.
Inventors: |
Pal; Biplab; (Ellicott City,
MD) ; Purohit; Amit; (Thane West, IN) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
PROPHECY SENSORS, LLC |
BALTIMORE |
MD |
US |
|
|
Family ID: |
57147600 |
Appl. No.: |
14/790084 |
Filed: |
July 2, 2015 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
14696402 |
Apr 25, 2015 |
|
|
|
14790084 |
|
|
|
|
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
G01D 7/005 20130101 |
International
Class: |
G01M 99/00 20060101
G01M099/00; G01D 7/00 20060101 G01D007/00 |
Claims
1. A method of a predictive maintenance IoT system comprising:
receiving, at a predictive maintenance IoT system, a plurality of
sensor data over a communications network; determining one or more
clusters from the sensor data based on a pre-determined rule set;
classifying the sensor data through a machine learning engine;
base-lining the sensor data through a combination of database
architecture, data training architecture, and a base-lining
algorithm; visualizing the machine state on a fuel gauge
representation based on a predictive maintenance state calculated
through a user regression model.
2. The method of claim 1, wherein the machine learning engine is
associated with at least one of a physics based model, a rule based
model and a vector classifier model.
3. The method of claim 1, wherein data training architecture
receives as input, one of a baseline reading and anomalous reading
for a component with a sensor attached.
4. The method of claim 1, wherein the fuel gauge is associated with
color schemes, wherein the color scheme: red indicates a worst
maintenance condition, yellow indicates an intermediate condition,
and green indicates a best maintenance condition.
5. The method of claim 1, wherein the sensor data is received from
a vacuum conveying system.
6. The method of claim 1, wherein a user defined alarm system is
set based on a field calibration of the fuel gauge
representation.
7. The method of claim 1, further comprising: raising an alarm when
color scheme is at least one of the yellow and the red; and
receiving sensor data from at least one machine wearable
sensor.
8. The method of claim 1, wherein the communications network is one
of WiFi, 2G, 3G, 4G, GPRS, EDGE, Bluetooth, ZigBee, Piconet of BLE,
Zwave, or a combination thereof.
9. The method of claim 1, wherein the sensor data is at least one
of a vibration, magnetic field, power factor, and temperature.
10. The method of claim 1, further comprising: wherein at least one
of a mobile, a web and a desktop application acts as a mobile
middleware, and wherein the mobile middleware calibrates and
base-lines the sensor data.
11. The method of claim 5, wherein the alarm is raised over the
communications network through one of a notification on the mobile
application, Short message service (SMS), email, or a combination
thereof.
12. A predictive maintenance IoT system comprising: a mobile
middleware to receive a plurality of sensor data over a
communications network; a real time data processing system
associated with distributed databases; a clustering module to
determine one or more clusters from the sensor data based on a
pre-determined rule set; a computer database to store the
pre-determined rule set; a machine learning engine to classify the
sensor data; a base-lining architecture to base-line the sensor
data, wherein the base-lining architecture is a combination of
database architecture, data training architecture, and a
base-lining algorithm; and a regression module associated with a
processor to predict a predictive maintenance state, wherein the
predictive maintenance state is mapped onto a depiction on a user
interface.
13. The system of claim 10, wherein the machine learning engine is
associated with at least one of a physics based model, a rule based
model and a vector classifier model.
14. The system of claim 10, wherein the data training architecture
receives as input, one of a baseline reading and anomalous reading
from a component associated with a sensor.
15. The system of claim 12, wherein the depiction on a user
interface is a fuel gauge, and wherein the fuel gauge is associated
with color schemes, wherein the color scheme: red indicates a worst
maintenance condition, yellow indicates an intermediate condition,
and green indicates a best maintenance condition.
16. The system of claim 12, further comprising: raising an alarm
when color scheme is at least one of the yellow and the red; and
receiving sensor data from at least one machine wearable
sensor.
17. The system of claim 12, wherein the communications network is
one of WiFi, 2G, 3G, 4G, GPRS, EDGE, Bluetooth, ZigBee, Piconet of
BLE, Zwave, or a combination thereof.
18. The system of claim 12, wherein the sensor data is at least one
of a vibration, magnetic field, power factor, and a
temperature.
19. The system of claim 12, wherein at least one of a mobile, a web
and a desktop application acts as a mobile middleware, and wherein
the mobile middleware calibrates and base-lines the sensor
data.
20. The system of claim 15, wherein the alarm is raised over the
communications network through one of a notification on the mobile
application, Short message service (SMS), email, or a combination
thereof.
Description
CROSS-REFERENCE TO RELATED APPLICATION
[0001] This application is a Continuation in Part of U.S. patent
application Ser. No. 14/696,402 filed Apr. 25, 2015, entitled
"INTERNET OF THINGS BASED DETERMINATION OF MACHINE RELIABILITY AND
AUTOMATED MAINTAINENACE, REPAIR AND OPERATION (MRO) LOGS", owned by
the assignees of the present application and herein incorporated by
reference in its entirety.
FIELD OF TECHNOLOGY
[0002] The present invention generally relates to Internet of
Things (IoT), more particularly relates to Predictive maintenance
through an IoT system based classification for pneumatic conveying
system.
BACKGROUND
[0003] Internet of Things (IoT) is a network of
uniquely-identifiable, purposed "things" that are enabled to
communicate data pertaining thereto, there between, over a
communications network whereby, the communicated data form a basis
for manipulating the operation of the "things". The "thing" in the
"Internet of Things" could virtually be anything that fits into a
common purpose thereof. For example, a "thing" could be a person
with a heart rate monitor implant, a farm animal with a biochip
transponder, an automobile that has built-in sensors to alert its
driver when tire pressure is low, or the like, or any other natural
or man-made entity that can be assigned a unique IP address and
provided with the ability to transfer data over a communication
network. Notably, if all the entities in an IoT are machines, then
the IoT is referred to as a "Machine to Machine" (M2M) IoT or
simply, as M2M IoT.
[0004] It is apparent from the aforementioned examples that an
entity becomes a "thing" of an M2M IoT especially, when the entity
is attached with one or more sensors capable of capturing one or
more types of data pertaining thereto: segregating the data (if
applicable); selectively communicating each segregation of data to
one or more fellow "things"; receiving one or more control commands
(or instructions) from one or more fellow "things" wherein, the one
or more control commands is based on the data received by the one
or more fellow "things"; and executing one or more commands
resulting in the manipulation or "management" of the operation of
the corresponding entity. Therefore, in an IoT-enabled system, the
"things" basically manage themselves without any human
intervention, thus drastically improving the efficiency
thereof.
[0005] US Patent publication 2014/0336791 A1 discusses a predictive
maintenance of industrial systems using big data analysis in a
cloud platform.
[0006] U.S. Pat. No. 8,560,368 B1 discusses constraint-based
scheduling, and in particular, constraint-based scheduling of one
or more components for maintenance based on both, time-based
maintenance information and condition-based maintenance
information.
[0007] U.S. Pat. No. 6,405,108 B1 discusses a system and process
for developing diagnostic algorithms for predicting impending
failures of the subsystems in a locomotive.
[0008] WIPO application WO2005086760 A2 discusses a method and
system, for monitoring and maintaining equipment and machinery, as
well as any other device or system that has discrete measuring
points that can be gathered and analyzed to determine the status of
the device or the system.
[0009] Visualization of analytical results or processed data from
big data system poses several new challenges in terms of
scalability, volume and velocity. Besides the results must be
interpreted to the users who are technicians and not familiar with
many of the advanced sensor data analytics. Therefore visualization
of the predictive maintenance results must be auto-interpreted to
factory technicians using simple normalized gauge scale concept.
None of the prior art technologies emphasize on the visualization
of the processed analytic data of predictive maintenance when
obtained as a result of complex machine learning calculation.
[0010] However, existing prior art technologies are limited to rule
based engines. Mere rule based engines do not provide effective
visualization of the equipment monitoring data which is critical
for operational deployment of predictive maintenance systems.
Further, mere rule based engines may not be sufficient to help
operators in handling multiple organ failure in machines. Further,
the above prior art technologies does not allow scalability in
order to handle large volumes of data and therefore not capable of
providing solution for TOT based predictive maintenance system.
[0011] It is evident from the discussion of the aforementioned
prior arts that none of the prior arts pave way for predictive
maintenance of a machine through an IoT system based classification
and providing effective visualization to a machine operator.
Therefore, there exists a need in the art for a solution to the
aforementioned problem.
SUMMARY
[0012] Disclosed are a method, an apparatus and/or a system of
predictive maintenance through an IoT system based
classification.
[0013] A method of predictive maintenance through an IoT system
comprises receiving a plurality of sensor data over a
communications network and determining one or more clusters from
the sensor data based on a pre-determined rule set. Further, the
sensor data is classified through a machine learning engine and the
sensor data is further base-lined through a combination of database
architecture, data training architecture, and a base-lining
algorithm. A predictive maintenance state is predicted through a
regression model and the predictive maintenance state is mapped
onto a depiction on a user interface.
[0014] A predictive maintenance based IoT system comprises: a
mobile middleware to receive a plurality of sensor data over a
communications network; a clustering module to determine one or
more clusters from the sensor data based on a pre-determined rule
set; a computer database to store the pre-determined rule set; a
machine learning engine to classify the sensor data; and a
base-lining architecture to base-line the sensor data. The
base-lining architecture is a combination of database architecture,
data training architecture and a base-lining algorithm. Further,
the system also includes a regression module associated with a
processor to predict a predictive maintenance state. The predictive
maintenance state is mapped onto a depiction on a user
interface.
[0015] The methods and systems disclosed herein may be implemented
in any means for achieving various aspects of intended results, and
may be executed in a form of a machine-readable medium embodying a
set of instructions that, when executed by a machine, cause the
machine to perform any of the operations disclosed herein. Other
features will be apparent from the accompanying drawings and from
the detailed description that follows.
BRIEF DESCRIPTION OF THE DRAWINGS
[0016] The embodiments of the present invention are illustrated by
way of example and not as limitation in the accompanying drawings,
in which like references indicate similar elements and in
which:
[0017] FIG. 1, is a diagrammatic representation of a predictive
maintenance IOT system, according to one or more embodiments.
[0018] FIG. 2, is a diagrammatic representation of a data
processing system capable of processing a set of instructions to
perform any one or more of the methodologies herein, according to
one embodiment.
[0019] FIG. 3, is a process flow diagram detailing the operations
of a method of a predictive maintenance IoT system, according to
one or more embodiments.
[0020] FIG. 4, is a diagrammatic representation of a fuel gauge to
depict a predictive maintenance state, according to one example
embodiment.
[0021] FIG. 5, is a flow diagram representing a
multi-classification, according to one embodiment.
[0022] Other features of the present embodiments will be apparent
from the accompanying drawings and from the detailed description
that follows.
DETAILED DESCRIPTION
[0023] Example embodiments, as described below, may be used to
provide a method, an apparatus and/or a system for predictive
maintenance through an IoT system based classification. Although
the present embodiments have been described with reference to
specific example embodiments, it will be evident that various
modifications and changes may be made to these embodiments without
departing from the broader spirit and scope of the various
embodiments.
[0024] One of the key emerging issues in IoT based systems is the
visualization of analytical results that have been obtained in real
time and/or near real time by processing data from
multiple-sensors. For a long time, sensor based systems remained
fairly simple consisting of only one kind of sensors. The one kind
of sensors indicated one of a failure and/or triggering event. The
IoT systems of the past were merely rule based alarms. Rules may
consist of a value higher and/or lower than a pre-assigned value.
With emerging IoT technologies, priority may be to extract useful
intelligence and/or meaningful information from data collected from
thousands of sensors of different kinds. Today, big data analytics
may allow for processing large data volumes at high speeds.
Therefore, a more complex alarm system may be designed wherein an
alarm threshold is not a simple sensor value but a complex
hyperplane constructed from a cluster of different kinds of sensor
values from different types of sensors (Ex: sensors with different
physical parameters such as temperature, pressure, vibration, power
factor etc.). Besides, sensor cluster data may consist of several
sensors, one single cluster may contain multiple alarm information
as multiple hyper-planes can be constructed for different kinds of
alarms and/or relevant information related to classification of
such cluster data.
[0025] However, one of the most prevailing issues of
multi-classification analytics is the effective visualization of
the processed result and/or alarm system in the specific case of
IoT. In a layered approach to sensor data analytics, classification
results may be mapped into a simple "Fuel Gauge" with a normalized
scale of 0-100, wherein a user can set a scale for setting up their
alarm and scaling up their predictive maintenance issue on the
field. Thus, complex results of Big Data IoT analytic may be
visualized in the most simplest and familiar form by applying the
techniques disclosed herein.
[0026] In an example embodiment, pneumatic conveying system
consists of vacuum pump, vacuum receiver, pickup device and
tubings. Vacuum pump may be the most critical equipment of the
vacuum conveying system.
[0027] In one or more embodiments, equipment may need to be
maintained at optimum condition to achieve efficient and smooth
performance. The equipment may undergo different failure modes
resulting in different types of faults. The type of fault
information may be critical for choosing maintenance actions for
the equipment. It is not possible to choose a maintenance action
without a knowledge of the fault type. On the other hand, incorrect
fault type information may lead to wasted maintenance effort and
subsequent equipment failure with safety and economic implications.
A Multi-fault classification technique may identify the fault type
of the equipment from an equipment operating data.
[0028] In an example embodiment, a vacuum conveying pump may
develop fault in different sub-components such as filter, oil, belt
blower, etc. More often different failure modes manifest in the
form of similar operator observation such as, deterioration in oil
quality and higher oil level resulting in increased vibration level
for the vacuum conveying pump. Therefore, it is very important to
detect the fault type correctly to plan the maintenance actions and
avoid safety related incidents and economic losses.
[0029] FIG. 1 is a system diagram of a predictive maintenance IoT
system, according to one or more embodiments. The predictive
maintenance IoT system 100 includes a machine 106, machine learning
engine 104, computer database 110, communications network 102, and
a mobile application 108.
[0030] In one or more embodiments, the predictive maintenance IoT
system 100 may comprise a mobile middleware 108 to receive a
plurality of sensor data from one or more machines such as machine
106 over a communications network 102. Each machine 106 may be
associated with a component and each component may be associated
with a machine wearable sensor. A clustering module may determine
one or more clusters from the sensor data based on a pre-determined
rule set stored in a computer database 110. A machine learning
engine 104 may classify the sensor data. Further, a base-lining
architecture may base-line the classified sensor data. The
base-lining architecture may be a combination of database
architecture, data training architecture, and a base-lining
algorithm. Further, the system may also include a regression module
associated with a computer processor to predict a predictive
maintenance state. The predictive maintenance state is mapped onto
a depiction on a user interface.
[0031] In an example embodiment, the predictive maintenance state
mapped onto a depiction on a user interface may be associated with
a mobile device running a mobile application 108.
[0032] In an example embodiment, the sensor data may be determined
from a machine wearable sensor placed on a motor, a machine
wearable sensor placed on the blower and so on.
[0033] In one or more embodiments, the communications network 102
may be one of a WiFi, 2G, 3G, 4G, GPRS, EDGE, Bluetooth, ZigBee,
Piconet of BLE, Zwave, or a combination thereof.
[0034] In one or more embodiments, the machine learning engine 104
may be associated with a machine learning algorithm. The machine
learning engine may be associated with one or more of one of a
physics based model, a rule based model and a vector classifier
model.
[0035] In one or more embodiments, a physics based model may
include extracting physical parameters from sensor data such as
total energy of vibration, multiple axes (X, Y , Z axis) of
vibration, azimuthal and polar angle of vibration rotation, RMS
(Root Mean Square) value of vibration, shape factor of vibration
and the like.
[0036] In one or more embodiments, the data training architecture
receives as input, one or more of a baseline reading and an
anomalous reading from a component associated with a sensor.
[0037] In an example embodiment, the depiction on a user interface
may be a fuel gauge type representation as shown in FIG. 4
conveying health monitoring system.
[0038] In one or more embodiments, the fuel gauge may be associated
with color schemes such as red, yellow and green. In the fuel gauge
color scheme, red may indicate a worst maintenance condition,
yellow may indicate an intermediate condition, and green may
indicate a best maintenance condition.
[0039] In one or more embodiments, an alarm may be raised when the
color scheme is one of a yellow and a red.
[0040] In one or more embodiments, the sensor data is one or more
of a vibration, magnetic field, power factor and a temperature.
[0041] In one or more embodiments, a mobile middleware is one of a
mobile application, a web application and a desktop application.
The mobile middleware may calibrate and base-line the sensor
data.
[0042] In one or more embodiments, base-lining may include adding a
data set automatically by running a good machine. Further, a good
machine may be a machine in an ideal state. The ideal state may be
a perception of a user of a machine. In predictive maintenance, the
basic objective is to find a difference between the good machine
and a bad machine. In older machines, the good machine may be a
machine that the user perceives as good. The good machine may not
have a new machine condition. Hence base-lining allows to take data
from machines in a learning mode to archive it as "data set" which
is referenced as "good machine condition", which is further used to
study deviation for bad machines.
[0043] In one or more embodiments, base-lining may be of different
types including manual and automatic. In manual base-lining, a user
may select any machine state and attribute the selected machine
state as a good baseline. In automatic base-lining, a factory
default machine data may be used as a reference and once the user
runs auto-baseline, a machine learning engine may adjust a factory
base-line level automatically.
[0044] In an example embodiment, multiple sensor data from multiple
locations may be received over a communications network 102 onto a
mobile application 108 coupled to a mobile device. An alarm may be
raised over the communications network 102 through one of a
notification on the mobile application, Short message service
(SMS), email, or a combination thereof.
[0045] FIG. 2 is a diagrammatic representation of a data processing
system capable of processing a set of instructions to perform any
one or more of the methodologies herein, according to an example
embodiment. FIG. 2 shows a diagrammatic representation of machine
in an exemplary form of a computer system 200 within which a set of
instructions, for causing the machine to perform one or more of the
methodologies discussed herein, may be executed. In various
embodiments, the machine operates as a standalone device and/or may
be connected (e.g., networked) to other machines.
[0046] In a networked deployment, the machine may operate in the
capacity of a server and/or a client machine in server-client
network environment, and or as a peer machine in a peer-to-peer (or
distributed) network environment. The machine may be a personal
computer (PC), a tablet PC, a set-top box (STB), a Personal Digital
Assistant (PDA), a cellular telephone, a web appliance, a network
router, switch and or bridge, an embedded system and/or any machine
capable of executing a set of instructions (sequential and/or
otherwise) that specify actions to be taken by that machine.
Further, while only a single machine is illustrated, the term
"machine" shall also be taken to include any collection of machines
that individually and/or jointly execute a set (or multiple sets)
of instructions to perform any one and/or more of the methodologies
discussed herein.
[0047] The example computer system 200 includes a processor 202
(e.g., a central processing unit (CPU) a graphics processing unit
(GPU) and/or both), a main memory 204 and a static memory 206,
which communicate with each other via a bus 208. The computer
system 200 may further include a video display unit 210 (e.g., a
liquid crystal displays (LCD) and/or a cathode ray tube (CRT)). The
computer system 200 also includes an alphanumeric input device 212
(e.g., a keyboard), a cursor control device 214 (e.g., a mouse), a
disk drive unit 216, a signal generation device 218 (e.g., a
speaker) and a network interface device 220.
[0048] The disk drive unit 216 includes a machine-readable medium
222 on which is stored one or more sets of instructions 224 (e.g.,
software) embodying any one or more of the methodologies and/or
functions described herein. The instructions 224 may also reside,
completely and/or at least partially, within the main memory 204
and/or within the processor 202 during execution thereof by the
computer system 200, the main memory 204 and the processor 202 also
constituting machine-readable media.
[0049] The instructions 224 may further be transmitted and/or
received over a network 226 via the network interface device 220.
While the machine-readable medium 222 is shown in an example
embodiment to be a single medium, the term "machine-readable
medium" should be taken to include a single medium and/or multiple
media (e.g., a centralized and/or distributed database, and/or
associated caches and servers) that store the one or more sets of
instructions. The term "machine-readable medium" shall also be
taken to include any medium that is capable of storing, encoding
and/or carrying a set of instructions for execution by the machine
and that cause the machine to perform any one or more of the
methodologies of the various embodiments. The term
"machine-readable medium" shall accordingly be taken to include,
but not be limited to, solid-state memories, optical and magnetic
media, and carrier wave signals.
[0050] FIG. 3 is a process flow diagram detailing the operations of
a method of a predictive maintenance through an IoT system,
according to one or more embodiments. A method of a predictive
maintenance through an IoT system may comprise the steps of:
receiving a plurality of sensor data over a communications network,
as shown in step 302 and determining one or more clusters from the
sensor data based on a pre-determined rule set, as shown in step
304. Further, the sensor data may be base-lined through a
combination of database architecture, data training architecture,
and a base-lining algorithm, as shown in step 306. Further, the
sensor data may be classified through a machine learning engine, as
shown in step 308. The intensity of fault state is mapped to a
depiction on a user interface, as shown in step 310 and the
predictive maintenance state is predicted through a regression
model, as shown in step 312.
[0051] FIG. 5 is flow diagram detailing the steps of a
multi-classification, according to one embodiment. In one or more
embodiments, steps of multi-classification may include data
transformation to achieve maximum separation among fault types, as
shown in step 502. Data transformation may lead to more accurate
multi classification e.g. linear discriminant functions. Further,
nonlinear hyper plane fitting may be done to classify different
fault types, as shown in step 504, e.g. quadratic hyper planes in
transformed variable space. Developing a measure to represent the
degree of fault based on machine learning multi-fault
classification approach. The intensity of fault may be calculated,
as shown in step 506, e.g. posterior probability of fault type. The
degree of fault information may be mapped to the fuel gauge, as
shown in step 508, e.g. combining different fault type posterior
probabilities to get fuel gauge representation. User calibration of
the fuel gauge, as shown in step 512 to include user intuition
about the machine state into the analytics process. The multi
classification may end when the user agrees with the fuel gauge, as
shown in step 510.
[0052] In an example embodiment, a predictive maintenance IoT
system may include machine wearable sensors. Further, the system
may be used for overseeing process control and predictive
maintenance of a machine or a network of machines. The system may
include a plurality of machine-wearable sensors, each of which may
be secured to the exterior of the machine. Each sensor is capable
of transmitting captured data wirelessly over a communications
network. The system may further include a sensor network for
receiving and transmitting the captured data over a communications
network and a machine learning algorithm engine capable of
receiving data from the sensor network. The machine learning
algorithm engine may process the received data to recognize one of
a pattern and a deviation, to issue control commands pertaining to
the machine. Lastly, the system may include one or more control
modules disposed in operative communication with a control panel of
the machine, the control module is capable of receiving, over a
communications network, one or more control commands and executing
the control commands.
[0053] In an example embodiment, the machine learning engine 104
may raise an alarm when one of a filter is clogged and deficient
oil is detected. The deficient oil may be one of a low oil level
and an overused oil structure. The system may be associated with a
plurality of machine wearable sensors. The machine learning engine
104 associated with the system, may issue commands based on a
learning outcome from an analysis through a combination of physics
based model, a rule based model and a vector classifier model of
the sensor data.
[0054] In one or more embodiments, the learning outcome may be
dependent on recognition of one of a pattern and deviation
recognized by the machine learning engine.
[0055] In one or more embodiments, the predictive maintenance IoT
system may detect multiple component or organ failure of a machine
before the failure happens.
[0056] In an example embodiment, temperature, vibration, power
factor and magnetic field data may be used for classification of a
machine state such as a bad oil level and/or low oil level. Bad oil
level may increase the friction and thereby raise the surface
temperature of a machine component.
[0057] In one or more embodiments, a machine learning based
classification may include a physics based classification, a vector
based classification, and a rule based classification. Vector based
classification may be based on an oblique and/or support vector
machine. Support vector machines may comprise supervised learning
models with associated learning algorithms that analyze data and
recognize patterns. The supervised learning models may be used for
classification and regression analysis.
[0058] In an example embodiment, a component such as blower may be
associated with machine wearable sensors. The machine wearable
sensors may measure multiple factors such as magnetic field,
surface temperature, and angular displacement.
[0059] In one or more embodiments, similar data patterns may be
found for multiple classes. A similar feature extraction may
indicate multiple root cause. For example higher RMS (Root mean
square) value of vibration may indicate multiple issues such as old
oil, high oil level, bearing failure etc.
[0060] In one or more embodiments, failure pattern data may be same
as some extreme process driven data. A process may generate a data
that may look like a failure data. Failure signs and/or failure
pattern data for different process conditions may be different.
[0061] In one or more embodiments, a multi-stage
multi-classification of IoT data for predictive maintenance of a
machine may include: a rule based process classification; building
clusters of sensor data based on models such as physics based
models; applying a multi-class classification to classify a sensor
data into various classes (such as bad oil, overfill oil etc., for
vacuum pump); discovering a root cause for a fault; and classifying
the sensor data into a predictive maintenance state such as a red,
yellow and green. Further, the classified sensor data may be mapped
onto a simple depiction such as a fuel gauge representation so that
a maintenance personnel may easily understand the
classification.
[0062] In one or more embodiments, a regression analysis may be a
statistical process for estimating a relationship amongst
variables. A regression algorithm may include fitting of a function
f(x, y, z) based on a scattered diagram of x, y, z, f etc. Storing
regression fitting parameters from a training data set may be
considered as simple machine learning algorithm.
[0063] In one or more embodiments, Principal component analysis
(PCA) may be a statistical procedure that uses an orthogonal
transformation to convert a set of observations of possibly
correlated variables into a set of values of linearly uncorrelated
variables called principal components. The number of principal
components selected for further analysis may be less than and/or
equal to the number of original variables. Projections may be made
onto a principal component subspace for a training data. Based on
the projections, classifications of hyper places may be made from
the training data.
[0064] In an example embodiment, data may be collected from diverse
locations such as 10,000 factory locations for 3P (prescriptive,
preventative and predictive) maintenance by using a combination of
Cassandra (distributed database), Storm and/or Spark real time to
process the data in a real time Big Data architecture and
implemented using a broker system such as Kafka for storing the
alarms as buffer database and then using Storm and/or Cassandra
like distributed database for an MRO (maintenance, repair and
operation) system.
[0065] In one or more embodiments, a prescriptive, preventative and
predictive maintenance may be a possibility for a machine. Big data
methodologies may be employed to analyze data obtained from various
locations through an IOT sensor network. Big data may be used to
describe a massive volume of both structured and unstructured data.
Large volumes of data may be difficult to process using a
traditional database and traditional software techniques.
Therefore, a distributed real-time computation system such as
Apache Storm may be used.
[0066] In an example embodiment, a real time data processing system
may be associated with distributed databases. The real time data
processing system may be a big data system.
[0067] FIG. 4 is an exemplary representation of a fuel gauge
depicting a predictive maintenance state, according to one
embodiment.
[0068] In an example embodiment, machines may run to failure very
often due to abusive operation coupled with poor maintenance.
Machines associated with machine wearable sensors may report one or
more sensor data such as temperature, vibration, pressure and
sound. These data may be used by a platform to check against a
baseline database and the platform offers early warning for machine
failure and/or real time alarm for faulty operation. From machine
learning algorithms of data, the platform sends out early
indication of machine failure and/or requirement of maintenance for
the machine.
[0069] In an example embodiment, machine learning of the
vibrational data may comprise transfer of vibrational energy from
one axis of rotation to other axis in order to determine an extent
of oldness of the oil, which is used in the blower bearings to
achieve smooth rotation. Machine learning of the vibrational data
may comprise of information related to instability and wobbling of
rigid rotational axis to determine an extent of oldness of oil used
in bearings of the blower.
[0070] In one or more embodiments, on field calibration of a fuel
gauge representation may be present, such that a maintenance
personal and/or user can set his/her own scale for setting up alarm
system and color scheme.
[0071] Although the present embodiments have been described with
reference to specific example embodiments, it will be evident that
various modifications and changes may be made to these embodiments
without departing from the broader spirit and scope of the various
embodiments. For example, the various devices and modules described
herein may be enabled and operated using hardware circuitry,
firmware, software or any combination of hardware, firmware, and
software (e.g., embodied in a machine readable medium). For
example, the various electrical structure and methods may be
embodied using transistors, logic gates, and electrical circuits
(e.g., application specific integrated (ASIC) circuitry and/or in
Digital Signal Processor (DSP) circuitry).
[0072] In addition, it will be appreciated that various operations,
processes and methods disclosed herein may be embodied in a
machine-readable medium and/or a machine accessible medium
compatible with a data processing system (e.g., a computer
devices), and may be performed in any order (e.g., including using
means for achieving the various operations). The medium may be, for
example, a memory, a transportable medium such as a CD, a DVD, a
Blu-ray.TM. disc, a floppy disk, or a diskette. A computer program
embodying the aspects of the exemplary embodiments may be loaded
onto a retail portal. The computer program is not limited to
specific embodiments discussed above, and may, for example, be
implemented in an operating system, an application program, a
foreground or background process, a driver, a network stack or any
combination thereof. The computer program may be executed on a
single computer processor or multiple computer processors.
[0073] Accordingly, the specification and drawings are to be
regarded in an illustrative rather than a restrictive sense.
* * * * *