• No results found

Implementation of Algorithm

In document DIAGNOSTICS OF INTERMITTENT ERRORS (Page 56-62)

The implementation of the algorithms developed in this master thesis has to be preceded by careful planning. Scania must decide which of the algorithms are to be implemented, which data storage method to use, and what threshold to deploy before any programming can occur.

Which algorithm to implement has to be determined to satisfy the first customer of the algorithm, namely the Workshop operator (as discussed in 3.1.1). If not, all algorithms developed in this master thesis are implemented, Scania should implement the algorithms recommended for each sensor. (see 5.1, 5.2, and 5.3)

The data storage method has to be determined to satisfy the second customer of the

algorithm, the R&D engineers at Scania. To satisfy the R&D engineers, Scania must retrieve IEs statistics (as discussed in 4.2.1 and 4.2.2). The statistics of interest are the progression of IEs over time and the IEs characteristics (as explained in 3.1.2). There are two ways of retrieving the statistics using the algorithms.

The first way is to use a logger device and deploy the matrix method of short-term memory storage. In the logger, the matrix is then continuously recorded. That way, excellent data with extremely high resolution of the IEs are obtained. However, it might be economically

unfeasible to deploy that many logger devices to get meaningful statistics.

The second way to retrieve the data is to continuously obtain the operational data from trucks that visit the workshop. That way, the algorithm can use the counter method, which saves slots in the short-term memory while still getting information regarding the progression of IEs. The data obtained by this method might be considered indistinct since usually there are extended periods in between workshop visits. However, it is economically feasible to retrieve statistics regarding IEs from a large pool of trucks using this method. Therefore the information is more statistically meaningful.

Scania can use both methods simultaneously; the first method on a small scale and the second method on a large scale. If both ways are deployed, Scania can perform detailed research based on the statistics from the first method. The fundamental analysis can be based on the statistics from the second method.

Scania must determine the threshold to satisfy the third customer of the algorithm, the legislative authorities, and vehicle safety. The threshold can be set using an arbitrary threshold and fine-tune the threshold along the way using data obtained during tests.

Alternatively, Scania can deploy the algorithms without a threshold at first. The algorithm values would then just be stored in the operational data. After some time, Scania can decide regarding thresholds using the operational data (as discussed in 4.3.4).

When Scania has decided what algorithms to use, the algorithm is to be coded. After the coding, the first step towards implementation is to do a vehicle test 1 (hereafter abbreviated to VT1).

The VT1 is performed on a small number of available test trucks at the Scania facility in Sodertalje, Sweden. The trucks with the new algorithms are then driven at a test track

designed to expose the trucks to various conditions that can occur when the customers drive the trucks. For example, the test track has steep hills, water obstacles, and sharp curves.

The purpose of the VT1 is to find out if the algorithm is working as intended. For example, if an unmotivated DTC is triggered, it suggests that the algorithm contains bugs that need to be addressed. Moreover, if some known hardware faults in the truck that the algorithm should identify but does not, it indicates that the algorithm does not work correctly. The VT1 is an iterative process. Usually, VT1’s are performed every ‘W’ week with updating and

improvements of the algorithm between every VT1 until the algorithm passes and is ready for the next step.

After the VT1, a vehicle test 2 (hereafter abbreviated to VT2) is performed. Scania has multiple customers with a large vehicle fleet that has volunteered to help Scania to perform tests. During the VT2, the algorithm is uploaded to a large number of vehicles and the performance of the new algorithm is monitored closely to evaluate its function.

If the algorithm passes the VT2, a decision-making meeting is had to discuss if the algorithm should be launched to production. Suppose it is decided at the meeting that the algorithm indeed should be established for production. In that case, it is sent to the production team at Scania, meaning that all new vehicles produced at Scania will have the new algorithm.

When the algorithm is launched, the department that has developed the algorithm writes a functionality specification. The functional specification aims to explain how the algorithm works and what it monitors. Moreover, it enables any employee at Scania to look up how the algorithm works. In some countries, a basic functionality specification must be filed with the national transport authority. For example, the requirements are very high in the USA, meaning that the transportation authority needs to know how the algorithm works to ensure that the truck meets the legislative requirements. In the EU, the conditions are not as steep;

however, a trend is seen that the requirements in the EU are going to increase to be more similar to the USA. [33]

Chapter 7: Conclusions

The possible symptoms in the sensor output signal when the sensor is suffering from an IEs are positive/negative peaks, oscillations, offset, damping, no signal, signal status

degradation, and maximum/minimum signal. Because of the various symptoms of IEs, multiple algorithms were developed which excel at capturing different signal behaviours. See Table 22for a summary of the best method for the respective IEs symptoms.

Table 22. Summary of what method is best for respective IEs.

IE signal symptom Best detection method

Short Peaks (One sample) Response Time Method / Out-of-Range Method Long Peaks (Multiple samples) Out-of-Range Method

Oscillations Spectral Analysis Method / Response Time

Method

No Signal or Signal Status Absent Signal Method / Signal Status Method Unfortunately, no algorithm was successfully developed for the detection of intermittent offsets and damping.

The methods with the lowest false positives are the out-of-range method, absent signal method, and signal status method. The response time method and the spectral analysis method can detect false positives because they can detect errors within range.

The best threshold value is still to be determined. There were too many unknown variables to calculate a theoretical threshold. Since no theoretical threshold was calculated, an arbitrary threshold should be set, which can be fine-tuned along the way. Alternatively, Scania can deploy the algorithms without a threshold at first. In that case, OBD would only store the values obtained in the operational data. After some time, Scania can use the operational data to set a threshold.

The fault isolation should be to the point of determining what measure to take at the

workshop. Therefore, the IEs should be categorized based on how the workshop operators can repair them.

Depending on what information is gathered, the storage solution for the detected IEs differs.

If a logger device is used, Scania should use the matrix method for enabling the logger to record high-resolution data. In all other cases, Scania should deploy the counter method. In all cases, the number stored in the operational data is one number per algorithm.

Scania should implement the algorithms and gather the data obtained (as proposed in 6.1).

Scania can then study the data to better understand IEs and fill the knowledge gap (as described in 2.6). Consequently, the increased knowledge can be used to further improve the algorithms for better detection of IEs and increase the overall performance of Scania vehicles.

References

[1] Wakil, “Intermittent fault finding strategies,” ScienceDirect, pp. 1-4, 2013.

[2] Avnet ABACUS, “www.avnet.com,” AVnet, [Online]. Available:

https://www.avnet.com/wps/portal/abacus/solutions/technologies/sensors/pressure-se nsors/core-technologies/capacitive/. [Accessed 20 04 2021].

[3] K. Hjort, “studentportalen.uu.se,” Lecture in Micro- and Nanotechnology, 2019. [Online].

[Accessed 20 04 2021].

[4] K. Reif, Diesel Engine Management, Weisbaden: Springer Vieweg, 2014.

[5] “Omega,” Omega Engineering Inc., 2003. [Online]. Available:

https://www.omega.com/en-us/resources/thermocouple-hub. [Accessed 07 06 2021].

[6] Integrated Device Technology, Inc., “www.renesas.com,” 2014. [Online]. Available:

https://www.renesas.com/eu/en/document/whp/tutorial-digital-sent-interface-zssc416x zssc417x?language=en. [Accessed 23 04 2021].

[7] Eclipseina GmbH, “youtube.com,” Eclipseina GmbH, 25 03 2019. [Online]. Available:

https://www.youtube.com/watch?v=VF1uUKLReFQ. [Accessed 22 04 2021].

[8] PHYS ORG., “NOx gases in diesel car fumes: Why are they so dangerous?,” Phys org, 09 2015. [Online]. Available:

https://phys.org/news/2015-09-nox-gases-diesel-car-fumes.html . [Accessed 26 02 2021].

[9] Diesel Net, “Heavy-Duty Truck and Bus Engines; Regulatory Framework,” [Online].

Available: https://dieselnet.com/standards/eu/hd.php . [Accessed 26 02 2021].

[10] C. Secrecy, Interviewee, R&D Engineer at Scania CV. [Interview]. 03 02 2021.

[11] Diesel Net, “dieselnet.com,” 2019. [Online]. Available:

https://dieselnet.com/tech/sensors_nox.php. [Accessed 21 01 2021].

[12] D. Secrecy, Interviewee, R&D Engineer at Scania CV. [Interview]. 23 04 2021.

[13] D. Atul, “Automotiveembeddedsite,” Automotive & Embedded Info. CAN & J1939, 2010.

[Online]. Available: https://automotiveembeddedsite.wordpress.com/can-j1939/.

[Accessed 23 02 2021].

[14] S. Khan, Philips and e. al, “No fault found, Retest ok, Cannot duplicate or fault not found?

Towards standardised taxonomy,” EPSRC Centre for Innovative Manufacturing in Through-life Engineering Services, pp. 3-6, 2015.

[15] A. Secrecy, Interviewee, Technical Manager at Scania SV. [Interview]. 26 02 2021.

[16] A. Correcher and E. García, “Diagnosis of Intermittent Faults and its dynamics, Factory Automation,” Javier Silvestre-Blanes(Eds.), 2010.

[17] A. Correcher, E. García, F. Morant, E. Quiles and L. Rodríguez, “Intermittent Failure Dynamics Characterization,” IEEE TRANSACTIONS ON RELIABILITY, vol. 61, no. 3, p. 650, 2012.

[18] P. Waszecki, M. Kauer, M. L. and S. Chakraborty, “Implicit Intermittent Fault Detection,”

IEEE, 2014.

[19] A. Erd and J. Stoklosa, “Failures of Electronic Systems and Elements in Electric Vehicles and Guidelines for Reducing Their Intensity,” Applications of Electromagnetics in Modern Engineering and Medicine (PTZE), pp. 38-41, 2019.

[20] A. Mahapatro and P. M. Khilar, “Fault Diagnosis in Body Sensor Networks,” International Journal of Computer Information Systems and Industrial Management Applications, vol. 5, pp.

252-259, 2012.

[21] Ubar, R. Raik, J. Vierhaus and H. Theodor, “Design and Test Technology for Dependable Systems-on-Chip,” IGI Global, 2011.

[22] Syed, W. Ahmad, Perinpanayagam, Suresh, Samie, Mohammad, Jennions and Ian, “A Novel Intermittent Fault Detection Algorithm and,” IEEE, vol. 6, no. 3, p. 400, 2016.

[23] J. Secrecy, Interviewee, Technical Manager at Scania CV. [Interview]. 23 03 2021.

[24] B. Secrecy, Interviewee, R&D Engineer at Scania CV. [Interview]. 08 04 2021.

[25] H. Secrecy, Interviewee, Senior Engineer at Scania CV. [Interview]. 12 05 2021.

[26] O. Secrecy, Interviewee, Development Engineer, Ph.D. [Interview]. 31 05 2021.

[27] R. Secrecy, Interviewee, R&D Engineer at Scania CV. [Interview]. 19 03 2021.

[28] V. Viswanathan, “mathworks.com,” MathWorks, 19 04 2017. [Online]. Available:

https://se.mathworks.com/videos/designing-and-implementing-real-time-signal-processi ng-systems-1502972724722.html. [Accessed 21 04 2021].

[29] M. Secrecy, “Internal Doc B,” Secrecy, Sodertalje, 2008.

[30] K. Secrecy, “Internal Doc A,” Secrecy, Sodertalje, 2021.

[31] N. Secrecy, “Internal Doc C,” Secrecy, Sodertalje, 2021.

[32] Universalclass, “www.universalclass.com,” Universal Class, [Online]. Available:

https://www.universalclass.com/articles/math/statistics/calculate-probabilities-normally -distributed-data.htm. [Accessed 29 06 2021].

[33] P. Secrecy, Interviewee, R&D Engineer at Scania CV. [Interview]. 26 04 2021.

[34] G. D'Agostini, “Probability and Measurement Uncertainty in Physics - A Bayesian Primer,”

University La Sapienza and INFN, Roma, Italy, vol. 1, no. 1, p. 34, 1995.

[35] B. Lindgren and V. Kapaklis, “Kompendium MÄTTEKNIK - Del 1,” Institution for physics and astronomy, p. 3, 2020.

Appendix

In document DIAGNOSTICS OF INTERMITTENT ERRORS (Page 56-62)

Related documents