As manufactured products become more complex, manufacturing equipment has increased in complexity at the system level. With the rise of complex manufacturing equipment, the time and monetary costs for new equipment has risen. This is compounded by untimely machine failure, which can halt production lines and increase the costs to a manufacturer. Due to this, maintenance is in a state of change between preventive maintenance and predictive maintenance. With predictive maintenance, however, there are challenges in how best to build models that provide worthwhile, accurate and understandable analysis to maintenance staff. Predictive Maintenance for larger systems, such as a vehicle skid, requires training data to “teach” algorithms about the different types of failures at the component and system level. This is typically carried out by instrumenting or sensing complex equipment behavior, waiting for a failure to occur, then examining the data to understand what signals to watch for in the future. This current process is tedious and sparse in the training data that it provides, as it can take months for a failure to be recorded, and in that instant captures only one known fault out of multiple potential failure modes. This paper proposes a method of creating accelerated failure data for rapidly building classifiers for predicting manufacturing components failure and then scale to larger system level manufacturing equipment. By incrementally introducing failures purposefully, the type of failure is controlled and known, and the development of overall systemic effects is observable and structurally classified. With knowledge of the lifetime of a component and failure types identified, the method provides trainable data to predict lifetime until failure and fault identification of that component. Using this method, a comprehensive understanding of the overall machine state is established with differentiation of failure at each component. The tested apparatus for the method is a bearing station using a small electric motor, with known bearing fault types introduced systematically, and vibration signals measured. After gathering the training data, a predictive classifier is built and tested against the failed components to validate the method’s training data in failure identification and prediction of remaining lifetime for a component. The method is scaled to system level equipment to validate the method’s potential of creating system level predictive models. Purposely failed components are installed on system level equipment to gather failure data while simulating a manufacturing process. A vehicle skid is used to validate the method by purposefully damaging components in a manner found in an automotive manufacturing facility.


A Proposed Method for Generating Lifetime Failure Data for Manufacturing Equipment: Validation With Bearings

Division / Sponsors: 
Aerospace, Bioengineering, Applied Mechanics, Design Engineering Division, Advanced Energy Systems, Manufacturing Engineering Division, Noise Control and Acoustics, Nondestructive Evaluation, Diagnosis, and Prognosis (NDPD), Nuclear Engineering Division, Process Industries Division, Fluids Engineering, Heat Transfer, Materials, Microelectromechanical Systems, Safety Engineering and Risk Analysis, Computers & Information in Engineering


Paper Proceedings: https://asmedigitalcollection.asme.org/IMECE/IMECE2020/volume/84478

IMECE2020 - Virtual Conference

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.