Diagnostic Features Modeling for Decision Boundaries Calculation for Maintenance of Gearboxes Used in Belt Conveyor System
Condition-Based maintenance (CBM) becomes more and more popular in industry. The idea is simple: measure raw data (vibrations, temperatures, etc.), extract features and make a right decisions regarding replacement of the whole machine or its component at appropriate time. Right decision might mean simple if-then-else rule or complex decision making scheme using multidimensional data. In any case mentioned rules require definition of appropriate thresholds for diagnostic parameters (i.e. decision boundaries). This is a key problem in CBM. The article presents the procedure for determining decision thresholds based on statistical modeling of diagnostic data. In the presented procedure first we fit the suitable distribution (Weibull) to data set for each gearbox. Next we calculate the fitting quality measure and select the distribution parameters for well fitted data. Finally, on the basis of the multidimensional analysis of those parameters we determine threshold values for the effective identification of the machines’ condition and their components. It might be interpreted as training process of diagnostic system. From this phase of the procedure we can obtain thresholds for warning and alarm statuses and they can be used for classification of rest of the data (that did not pass modeling phase). Proposed procedure has been applied to relatively large diagnostic data set that covers nearly 150 measurements acquired during several years in underground mine. The data describes gearboxes in different conditions—from nearly new or after repair to seriously damaged/worn just before failure.
KeywordsBelt conveyor Decision making Maintenance Weibull distribution Diagnostics
This work is partially supported by the statutory grant No. B40037 (P. Stefaniak).
- 3.Bartelmus W (1992) Mach Vibr 1:178–189Google Scholar
- 8.Coles S (2001) An introduction to statistical modeling of extreme values. Springer, BerlinGoogle Scholar
- 11.Galar D, Gustafson A, Tormos B, Berges L (2012) Maintenance decision making based on different types of data fusion. Eksploatacja I Niezawodnosc—Maintenance Reliab 14(2):135–144Google Scholar
- 13.Lodewijks G (2004) Strategies for automated maintenance of belt conveyor systems. Bulk Solids Handling 24(1):16–22Google Scholar
- 14.Stefaniak PK, Wyłomańska A, Obuchowski J, Zimroz R (2015) Procedures for decision thresholds finding in maintenance management of belt conveyor system—statistical modeling of diagnostic data. Proceedings of the 12th international symposium continuous surface mining—Aachen 2014. Lecture notes in production engineering 2015, pp 391–402Google Scholar
- 19.Zimroz R, Stefaniak P, Bartelmus W, Hardygóra M (2015) Novel techniques of diagnostic data processing for belt conveyor maintenance. Proceedings of the 12th international symposium continuous surface mining—Aachen 2014. Lecture notes in production engineering 2015, pp 31–40Google Scholar
- 20.Lim CKR, Mba D (2014) Switching Kalman filter for failure prognostic. Mech Syst Sig Process. http://dx.doi.org/10.1016/j.ymssp.2014.08.006i
- 22.Stefaniak PK, Zimroz R, Krol R, Gorniak-Zimroz J, Bartelmus W, Hardygora M (2012) Some remarks on using condition monitoring for spatially distributed mechanical system belt conveyor network in underground mine—a case study condition monitoring of machinery in non-stationary operations: proceedings of the second international conference. Condition monitoring of machinery in non-stationary operations. CMMNO’ 2012/ed. TaharFakhfakh [i in.] Springer, pp 497–507Google Scholar