Abstract: Supervised learning is implemented to improve the accuracy of automated diagnoses performed by monitoring units installed at a machine. The monitoring units perform indicator acquisition and automated diagnoses based on a Bayesian model derived in accordance with the machine's known configuration. Raw data is collected, including machine vibration data and other diagnostic data. The data is analyzed to diagnose for specific fault defect assumptions so as to generate the automated diagnoses results and a rating for overall health of the machine. The results are uploaded to an external environment that can be accessed by an expert for review and correction. Based upon the expert's corrections, the Bayesian model is adjusted using supervised learning to improve the automated diagnoses performed by the monitoring units.

Abstract: Supervised learning is implemented to improve the accuracy of automated diagnoses performed by monitoring units installed at a machine. The monitoring units perform indicator acquisition and automated diagnoses based on a Bayesian model derived in accordance with the machine's known configuration. Raw data is collected, including machine vibration data and other diagnostic data. The data is analyzed to diagnose for specific fault defect assumptions so as to generate the automated diagnoses results and a rating for overall health of the machine. The results are uploaded to an external environment that can be accessed by an expert for review and correction. Based upon the expert's corrections, the Bayesian model is adjusted using supervised learning to improve the automated diagnoses performed by the monitoring units.

Abstract: Supervised learning is implemented to improve the accuracy of automated diagnoses performed by monitoring units installed at a machine. The monitoring units perform indicator acquisition and automated diagnoses based on a Bayesian model derived in accordance with the machine's known configuration. Raw data is collected, including machine vibration data and other diagnostic data. The data is analyzed to diagnose for specific fault defect assumptions so as to generate the automated diagnoses results and a rating for overall health of the machine. The results are uploaded to an external environment that can be accessed by an expert for review and correction. Based upon the expert's corrections, the Bayesian model is adjusted using supervised learning to improve the automated diagnoses performed by the monitoring units.

Abstract: A wireless monitoring unit performs condition monitoring associated with a test point location of a host machine. The unit includes a high performance vibration sensor useful for monitoring the host machine for fault conditions and the like. The unit includes a second vibration sensor that has poorer metrological performance, but draws a comparatively insignificant amount of power from the battery. The second sensor takes a vibration measure periodically, which measure is processed to evaluate whether the machine is in a state eligible to undergo a health assessment. Diagnostic monitoring using the high performance vibration sensor is enabled only when the host machine it is in the eligible state. A low power temperature sensor provides additional data for guarding against a false trigger, in effect allowing the processor to distinguish against vibration from another machine propagating to the host machine causing the host machine to vibrate while the host machine is not running.

Abstract: Automatic fault diagnosis is performed on vibration data sensed from a machine. A set of faults to screen for is identified from the machine configuration. For each fault there are characteristic symptoms. For each characteristic symptom, there is a corresponding indication used to diagnose the symptom. The indications are based on analyses of the current vibration data. The diagnosed symptoms have weights assigned according to a Bayesian network, and are used to derive a Bayesian probability for the fault. A fault having a Bayesian probability exceeding a threshold value is identified as being present in the machine. For each fault a confidence level is derived. The confidence level for a first fault is based on a similarity between characteristic symptoms for the first fault and characteristic symptoms for each one of the other faults being screened.

Abstract: A wireless monitoring unit performs condition monitoring associated with a test point location of a host machine. The unit includes a high performance vibration sensor useful for monitoring the host machine for fault conditions and the like. The unit includes a second vibration sensor that has poorer metrological performance, but draws a comparatively insignificant amount of power from the battery. The second sensor takes a vibration measure periodically, which measure is processed to evaluate whether the machine is in a state eligible to undergo a health assessment. Diagnostic monitoring using the high performance vibration sensor is enabled only when the host machine it is in the eligible state. A low power temperature sensor provides additional data for guarding against a false trigger, in effect allowing the processor to distinguish against vibration from another machine propagating to the host machine causing the host machine to vibrate while the host machine is not running.

Abstract: A sensor unit is configured as a single body, removably mounted in its entirety to a test point location on a machine so machine vibrations propagate into the single body. Within are an accelerometer, circuit board, wireless interface, signal processor, and battery. The sensor unit transmits sensor data wirelessly in real time to a data collection unit. A technician with data collection unit in hand goes from machine to machine, along a route of multiple test point locations on multiple machines, mounting and dismounting the sensor unit and collecting machine vibration data. The sensor unit is configured to reduce frequency response impacts of the mass and volume of the circuit board, wireless interface, signal processor, and battery on dynamic behavior of the sensor unit with respect to machine vibrations to achieve a frequency response rating comparable to a wired sensor.

Abstract: Data acquisition and automatic diagnosis is performed at test points by respective wireless monitoring units (WMU). Each WMU tests a first indicator periodically at a first rate to determine whether to test a full set of indicators. Evolution of the first indicator determines the pass/fail result of a reduced assessment of the corresponding test point. The pass/fail result is automatically sent off-site to a central database. When a fail occurs, a full assessment is performed, which includes monitoring the full set of indicators. Automatic diagnosis is performed based on the full set of indicators. The reduced assessment does not include automatic diagnosis of a test point, but is used to decide whether to perform the full assessment and automatic diagnosis. Results of the automatic diagnosis are sent to the central database. Raw data is not sent automatically.

Abstract: A sensor unit is configured as a single body, removably mounted in its entirety to a test point location on a machine so machine vibrations propagate into the single body. Within are an accelerometer, circuit board, wireless interface, signal processor, and battery. The sensor unit transmits sensor data wirelessly in real time to a data collection unit. A technician with data collection unit in hand goes from machine to machine, along a route of multiple test point locations on multiple machines, mounting and dismounting the sensor unit and collecting machine vibration data. The sensor unit is configured to reduce frequency response impacts of the mass and volume of the circuit board, wireless interface, signal processor, and battery on dynamic behavior of the sensor unit with respect to machine vibrations to achieve a frequency response rating comparable to a wired sensor.

Abstract: Automatic fault diagnosis is performed on vibration data sensed from a machine. A set of faults to screen for is identified from the machine configuration. For each fault there are characteristic symptoms. For each characteristic symptom, there is a corresponding indication used to diagnose the symptom. The indications are based on analyses of the current vibration data. The diagnosed symptoms have weights assigned according to a Bayesian network, and are used to derive a Bayesian probability for the fault. A fault having a Bayesian probability exceeding a threshold value is identified as being present in the machine. For each fault a confidence level is derived. The confidence level for a first fault is based on a similarity between characteristic symptoms for the first fault and characteristic symptoms for each one of the other faults being screened.