Blog Post

DOCTORAL THESIS BY DR CESAR SOTO ON CONDITION BASED MONITORING, AWARDED SUMMA CUM LAUDE HONOURS

8 de mayo de 2023

Doctoral Thesis by Cesar Soto Ocampo, Development of a monitoring system for rotating machinery – Application to railway axle support bearings, with which he has been granted the title of PhD in Industrial Engineering with the maximum qualification of Summa Cum Laude.



CITEF’s greatest asset consists of its human team, proof of which is this Doctoral Thesis by Cesar Soto Ocampo, Development of a monitoring system for rotating machinery – Application to railway axle support bearings, with which he has been granted the title of PhD in Industrial Engineering with the maximum qualification of Summa Cum Laude. The research for this thesis was carried out at CITEF under the supervision of Professor José Manuel Mera Sánchez de Pedro.


This doctoral thesis researches the implementation of condition monitoring strategies to assess the condition of railway grease box bearings. The main objective is to use early onset parameters, such as vibration signals, which allow identifying the presence of anomalies early on. This allows developing a maintenance plan adapted to the condition of each element, thus reducing action costs and times, as well as unexpected downtime and failure-related risks and costs.


An initial literature review revealed that the main barriers limiting the implementation of condition monitoring strategies (based on vibration signals) lie on the high cost of the different sensing equipment and the complexity of the different diagnostic methods. In view of this, the thesis focuses in overcoming these difficulties by developing, on the one hand, a data acquisition system based on low cost and high performance components and, on the other hand, an intuitive and easily implemented diagnostic method based on contour maps.


The research carried out for this thesis has produced the following publications:


References

  • Soto-Ocampo, C. R., Mera, J. M., Cano-Moreno, J. D., & Garcia-Bernardo, J. L. (2020). Low-Cost, High-Frequency, Data Acquisition System for Condition Monitoring of Rotating Machinery through Vibration Analysis-Case Study. Sensors, 20(12), Article 12. https://doi.org/10.3390/s20123493. SJR 0.8 (Q1)
  • Soto-Ocampo, C. R., Cano-Moreno, J. D., Mera, J. M., & Maroto, J. (2021a). Bearing Severity Fault Evaluation Using Contour Maps—Case Study. Applied Sciences, 11(14), Article 14. https://doi.org/10.3390/app11146452. SJR 0.51 (Q2)

International Conventions 

  • Soto-Ocampo, C. R., Cano Moreno, J. D., & Mera, J. M. (2022). Construction of fault evolution maps for railway bearing damage level assessment. 213, 67-79. https://doi.org/10.2495/CR140351
  • Soto-Ocampo, C. R., Cano-Moreno, J. D., Mera, J. M., & Gil González, R. (2022). Método de diagnóstico de fallas de rodamientos basado en mapas de contorno. 2, 172-183. https://doi.org/10.5944/bicim2022.333

Databases:

  • Soto-Ocampo, C., Mera, J., Cano-Moreno, J., & Garcia, José. (2020). Bearing Database, Induced Failure Test Data on Rolling Elements of a Spherical Roller Bearing (FAG 22205E1KC3) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.3898942
  • Soto-Ocampo, C. R., Cano-Moreno, J. D., Mera, J. M., & Maroto, J. (2021b). Bearing Database—Combined Failure, Failure-induced vibration recordings on outer race and rolling elements of a spherical roller bearing (FAG 22205E1KC3) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.5084405

The work described in this thesis is presented below: 


Data acquisition system (DAQ)


The DAQ has been developed taking into account the essential requirements for analysing bearing condition using vibration analysis. One of the main objectives is to provide instruments for the four grease boxes of a bogie, which requires at least four recording channels. We have also provided for 16-bit resolution, comparable to that of commercial equipment. We have also considered the sampling frequency required by the equipment to acquire vibration signals from bearings, which is 50 kHz in each channel. And, finally, that it is a full DAQ module including a CPU unit and a converter card.


To meet these requirements, a single-board Raspberry Pi 3 B+ (RPi) computer was used as the CPU unit, which provides the following advantages: it has an independent operating system, it is compatible with several programming languages, the data is stored in the RPi’s internal memory and, most interestingly, it allows remote connections. 

The converter card has been provided with four recording channels linked to a multiplexor, which will command the connection of each connected channel and store the data until the 16-bit resolution converter has completed its task. It is worth highlighting that the software that controls and manages data capture is installed in the RPi memory, and therefore this control can be performed remotely.


The operation of this DAQ has been tested by performing a few validation tasks, both capturing known wave signals and vibration signals associated to different bearing conditions. In each of these cases there was great data capture performance, both in identifying the known wave frequency and evaluating the frequency and classical failure modes of the bearing components being evaluated. These tests have allowed us to determine the maximum work performance, reaching a sampling frequency of up to 200 kHz. This demonstrates the possibility of developing low-cost, high-performance data acquisition systems, as well as the viability of implementing them in bearing condition monitoring tasks. The main features of the DAQ developed in this study are listed below:



 

  • 16-bit resolution, 
  • 5V power supply,
  • 4 recording channels,
  • High sampling frequency in each channel, 
  • Direct data storage,
  • Data capture controlled remotely,
  • Portable and easy-to-implement equipment,
  • Multipurpose equipment,
  • Low cost. 

Diagnostic methodology


This methodology is based on the following concepts:


  • Oscillation foundations of bearings. Bearings have specific vibration signatures created by the interaction of their different components. This establishes a relationship between the state of the surface of said components and the amplitude of their working frequency.
  • Constructive features of contour maps. These graphs have been widely used to characterise environments at different elevations (such as in topography, meteorology, navigation, etc.) by using isolines. This allows us to establish a relationship between the three variables that describe said environment or space.



This thesis uses contour maps to characterise different records that describe the behaviour of the bearing over time (see Figura 2), establishing a relationship between failure frequency, failure amplitude and the level of failure the bearing is in. This allows assessing the future condition of the bearing based on the characterisation of previously calculated failure evolution maps. This methodology has also been complemented with predictive models based on multiple linear regressions, making it possible to define the failure level of the bearing and the magnitude of said failure.



This methodology has been validated in two phases. We first built respective failure evolution maps, estimated the prediction model and, finally, performed diagnostic accuracy tests using records from a first test. This has allowed us to reach a diagnostic accuracy of 91% in the study case.


Finally, these failure evolution maps and predictive models were used to assess records from a second test with different failure magnitudes. This has shown a great performance of the methodology regarding diagnosis, demonstrating that contour maps are a powerful tool to characterise the behaviour of bearing components and that when linked to predictive models a relationship can be established not only with the level of failure but also with the area affected. One of the most representative properties of the methodology is the reliability of the diagnosis provided from records between optimal working conditions and the presence of anomalies, where there have been no false positives or false negatives in any of the cases analysed. In addition, since this is a graphical technique it is easy to interpret and does not demand a high degree of training from the analyst.


Share by: