REAL TIME MACHINE LEARNING BASED PREDICTIVE AND PREVENTIVE MAINTENANCE OF VACUUM PUMP

• US 20160245279A1
• Filed: 02/23/2015
• Published: 08/25/2016
• Est. Priority Date: 02/23/2015
• Status: Abandoned Application

• Current Assignee: Machinesense LLC
• Original Assignee: Machinesense LLC
• Inventors: Gillmeister, Steve, Purohit, Amit, Pal, Biplab
• Application Number:
  US14/628,322
• Publication Number:
  US 20160245279A1
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: F04B 51/00 Testing machines, pumps, or pumping installations
• CPC Class Codes: G01M 3/025 Details with respect to the testing of engines or engine parts
• CPC Class Codes: G01M 3/26 by measuring rate of loss or gain of fluid, e.g. by pressure-responsive devices, by flow detectors

Abstract

A method and system of a machine learning architecture for predictive and preventive maintenance of vacuum pumps. The method includes receiving one of a motor sensor data and a blower sensor data over a communications network. The motor sensor data is classified into one of a vacuum state sensor data and break state sensor data. The vacuum state sensor data is analyzed to detect an operating vacuum level and an alarm is raised when the vacuum state sensor data exceeds a pre-defined safety range. Vacuum break data is classified into one of a clean filter category and clogged filter category and an alarm is raised if an entry under the clogged filter category is detected. The blower sensor data in association with the motor sensor data is analyzed based on machine learning to detect one of a deficient oil level and a deficient oil structure.

Claims

1. A method of a machine learning architecture comprising:

i) receiving at least one of a motor sensor data and a blower sensor data over a communications network,

wherein one of the motor sensor data and the blower sensor data comprises at least one of a vibration, a magnetometer, a gyroscope, a sound and a temperature;

ii) classifying at least one of the motor sensor data and the blower sensor data into one of a vacuum state sensor data and break state sensor data,

wherein at least one of the motor sensor data and the blower sensor data are classified by one of individually and in combination,

wherein the break state sensor data is received when a rotor of a vacuum pump is malfunctioning;

iii) analyzing the vibration data of the vacuum state sensor data to detect an operating vacuum level,

wherein an alarm is raised when the vacuum state sensor data of one of a vibration and a temperature exceeds a pre-defined safety range; and

iv) classifying vacuum break data into one of a clean filter category and clogged filter category,

wherein the alarm is raised if an entry under the clogged filter category is detected; and

analyzing the blower sensor data in association with the motor sensor data based on machine learning to detect at least one of a deficient oil level and a deficient oil structure.

2. The method of claim 1, further comprising:

receiving sensor data from at least one machine wearable sensor placed on one of a motor and a blower.

3. The method of claim 1, wherein the communications network between sensor and the data collection unit, comprises one of WiFi, 2G, 3G, 4G, GPRS, EDGE, Bluetooth, ZigBee, Piconet of BLE, Zwave, or a combination thereof, wherein the data collection unit is one of a mobile device and a wireless enabled device.

4. The method of claim 1, further comprising:

associating the machine learning architecture with a machine learning algorithm where, normal states of the vacuum pumps with operational range, clean filer and clean oil are learned with a baseline reading and anomalous readings from one of a clogged filter, a bad operation, a bad oil, a low oil level and an over filled oil level, wherein the baseline reading and the anomalous readings are used as a training database.

5. The method of claim 1, further comprising:

acquiring data from multiple vacuum pumps associated with machine wearable sensors,

wherein at least one of a mobile, a web and a desktop application acts as a mobile middleware to scale the machine learning architecture to a single data collection unit, and

wherein the single data collection unit is at least one of a mobile device and a wireless device.

6. The method of claim 1, 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.

7. A machine learning architecture comprising:

a vacuum pump including a motor and a blower;

a motor associated with a machine wearable sensor;

a blower associated with a another machine wearable sensor,

wherein at least one of a motor sensor data and a blower sensor data is received over a communications network,

wherein at least one of the motor sensor data and the blower sensor data is at least one of a vibration, a sound and a temperature,

wherein at least one of the motor sensor data and the blower sensor data is classified into one of a vacuum state sensor data and break state sensor data,

wherein at least one of the motor sensor data and the blower sensor data are classified by one of individually and in combination,

wherein the break state sensor data is received when a rotor of a vacuum pump is malfunctioning,

wherein the vibration data of the vacuum state sensor data is analyzed to detect an operating vacuum level,

wherein an alarm is raised when the vacuum state sensor data of one of a vibration and a temperature exceeds a pre-defined safety range,

wherein vacuum break data is classified into one of a clean filter category and clogged filter category,

wherein the alarm is raised if an entry under the clogged filter category is detected; and

wherein the blower sensor data is analyzed in association with the motor sensor data based on machine learning to detect at least one of a deficient oil level and a deficient oil structure.

8. The architecture of claim 7, wherein machine learning is used on a vibrational data transformed by PCA (Principal component analysis) transformation of X, Y and Z axis components of the vibrational data to transpose the acceleration into reference frame of the rotor of the vacuum pump.

9. The architecture of claim 7, wherein machine learning of the vibrational data which comprises transfer of vibrational energy from one axis of rotation to other axis to determine extent of oldness of the oil used in the blower bearings for smooth rotation.

10. The architecture of claim 7, wherein machine learning of the vibrational data comprises 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.

11. The architecture in claim 7, wherein machine learning of the vibrational data comprises of information related to shape factor of the vibration calculated as a ratio of moving RMS value to moving average of absolute value.

12. A predictive and preventive maintenance system for a vacuum pump comprising:

one or more machine wearable sensors associated with the vacuum pump;

a tracking module associated with a computing device;

a machine learning module associated with a database; and

a communications network,

wherein at least one changing condition of vacuum pump is tracked through the tracking module over the communications network,

wherein the tracking module receives at least one of a temperature, a vibration and a sound data from the one or more machine wearable sensors,

wherein the machine learning module associated with the tracking module identifies a pattern from the temperature, the sound and the vibration data, and

wherein the machine learning module raises an alarm based on an analysis of the pattern.

13. The system of claim 12, wherein the machine learning algorithm engine raises an alarm when at least one of a filter is clogged and deficient oil is detected.

14. The system of claim 12, wherein the deficient oil is one of a low oil level and an overused oil structure.

15. The system of claim 12, wherein the one or more machine wearable sensors include at least one of a motor sensor and a blower sensor.

16. The system of claim 12, wherein the machine learning module engine is associated with an IoT based system,

wherein the machine learning module issues commands based on a learning outcome from the at least one changing condition.

17. The system of claim 16, wherein the learning outcome is dependent on recognition of at least one of a pattern and deviation by the machine learning module.