Automated machine analysis
A method for automated condition monitoring whereby techniques of automated vibration analysis and signal processing are combined with deep learning/machine learning techniques for an enhanced system of automated anomaly detection, problem classification, and problem regression. The method may be implemented in software, firmware or hardware to run autonomously. Machines monitored and analyzed according to the disclosed method are typically found in industrial plants or commercial applications, but the disclosed invention may be applied to any rotating equipment such as motors, fans, pumps, compressors, and etc., in any environment where they are functioning.
1. A method for determining wear conditions in a machine that produces vibration of mechanical parts of said machine, said method comprising the steps of:
generating signals from at least one sensor that produces signals in response to vibration in said mechanical parts of said machine;
storing data from signals that are generated by said at least one sensor that are generated in response to vibrations of said mechanical parts of said machine;
determining at least one synchronous harmonic family within data from signals that are generated by said at least one sensor, wherein each of said at least one synchronous harmonic family includes one or more member signals that represent a fundamental frequency or an integral multiples of said fundamental frequency,
determining at least one non-synchronous harmonic family within data from signals that are generated by said at least one sensor, wherein each of said at least one non-synchronous harmonic family includes one or more member signals having a frequency that is not an integral multiple of said fundamental frequency;
associating the at least one non-synchronous harmonic family with a specific mechanical part of said machine;
identifying sidebands that correspond to each member of said at least one synchronous harmonic family or to each member of said at least one non-harmonic family, wherein said sidebands of said member of said synchronous harmonic family or said sidebands of each member of said non-harmonic family are identified according to either amplitude modulation or frequency modulation or both amplitude and frequency modulation in comparison to the member signals of said at least one synchronous harmonic family or in comparison to the member signals of said at least one non-synchronous harmonic family; and
identifying trends in the members of said synchronous harmonic families, the members of said non-synchronous harmonic families, and/or the sidebands of said members of said synchronous harmonic families or the sidebands of said member of said non-synchronous harmonic families that correspond to abnormal wear or abnormal operation of said mechanical parts of said machine.
2. The method of claim 1 wherein said step of identifying trends includes comparing the amplitude of said members of said non-synchronous harmonic families, and/or the sidebands of said members of said synchronous harmonic families, and/or the sidebands of said members of said non-synchronous harmonic families to the progression of mechanical wear of said mechanical parts of said machine.
3. The method of claim 1 wherein said step of identifying trends includes comparing the frequency of the members of said non-synchronous harmonic families, and/or the sidebands of said members of said synchronous harmonic families or the sidebands of said members of said non-synchronous harmonic families to the progression of mechanical wear of said mechanical parts of said machine.
4. The method of claim 1 wherein said step of identifying trends includes comparing two or more members of said synchronous harmonic family to the progression of mechanical wear of said mechanical parts of said machine.
5. The method of claim 1 wherein said step of identifying trends includes comparing two or more members of said non-synchronous harmonic family to the progression of mechanical wear of said mechanical parts of said machine.
6. The method of claim 1 wherein said step of identifying trends includes comparing two or more sidebands of members of said synchronous harmonic family to the progression of mechanical wear of said mechanical parts of said machine.
7. The method of claim 1 wherein said step of identifying trends includes comparing two or more sidebands of members of said non-synchronous harmonic family to the progression of mechanical wear of said mechanical parts of said machine.
8. The method of claim 1 wherein said step of identifying trends includes comparing two or more sidebands of members of two or more non-synchronous harmonic families to the progression of mechanical wear of said mechanical parts of said machine.
9. The method of claim 1 wherein said synchronous harmonic family is determined according to the total energy of the members of said synchronous harmonic family.
10. The method of claim 1 wherein said non-synchronous harmonic family is determined according to the total energy of the members of said non-synchronous harmonic family.