Development of a methodology for the diagnosis of internal combustion engines using non-invasive measurements based on the use of interpretable neural networks applicable to databases with multiple annotators

dc.contributorGil González, Julián
dc.creatorValencia Duque, Andrés Felipe
dc.date2022-11-10T14:58:49Z
dc.date2022-11-10T14:58:49Z
dc.date2022
dc.date.accessioned2023-06-05T15:17:55Z
dc.date.available2023-06-05T15:17:55Z
dc.identifierUniversidad Tecnológica de Pereira
dc.identifierRepositorio Institucional Universidad Tecnológica de Pereira
dc.identifierhttps://repositorio.utp.edu.co/home
dc.identifierhttps://hdl.handle.net/11059/14367
dc.identifier.urihttps://repositorioslatinoamericanos.uchile.cl/handle/2250/6645779
dc.descriptionPressure is one of the essential variables that give information for engine condition and monitoring. Direct recording of this signal is complex and invasive, while the angular velocity can be measured easily. Nonetheless, the challenge is to predict the cylinder pressure using the shaft kinematics accurately. On the other hand, the increasing popularity of crowdsourcing platforms, i.e., Amazon Mechanical Turk, changes how datasets for supervised learning are built. In these cases, instead of having datasets labeled by one source (which is supposed to be an expert who provided the absolute gold standard), databases holding multiple annotators are provided. However, most state-of-the-art methods devoted to learning from multiple experts assume that the labeler's behavior is homogeneous across the input feature space. Besides, independence constraints are imposed on annotators' outputs. This document presents a Regularized Chained Deep Neural Network to deal with classification tasks from multiple annotators. In this thesis, we develop 2 strategies aiming to avoid intrusive techniques that are commonly used to diagnose Internal Combustion Engines (ICE). The first consist of a time-delay neural network (TDNN), interpreted as a finite pulse response (FIR) filter to estimate the in-cylinder pressure of a single-cylinder ICE from fluctuations in shaft angular velocity. The experiments are conducted over data obtained from an ICE operating in 12 different states by changing the angular velocity and load. The TDNN's delay is adjusted to get the highest possible correlation-based score. Our methodology can predict pressure with an R2>0.9, avoiding complicated pre-processing steps. The second technique, termed RCDNN, jointly predicts the ground truth label and the annotators' performance from input space samples. In turn, RCDNN codes interdependencies among the experts by analyzing the layers' weights and includes l1, l2, and Monte-Carlo Dropout-based regularizers to deal with the overfitting issue in deep learning models. Obtained results (using both simulated and real-world annotators) demonstrate that RCDNN can deal with multi-labelers scenarios for classification tasks, defeating state-of-the-art techniques.
dc.descriptionLa presión es una de las variables esenciales que dan información para el estado del motor y su monitorización. El registro directo de esta señal es complejo e invasivo, mientras que la velocidad angular puede medirse fácilmente. No obstante, el reto consiste en predecir la presión del cilindro utilizando la cinemática del eje con precisión. Por otro lado, la creciente popularidad de las plataformas de crowdsourcing, por ejemplo, Amazon Mechanical Turk, cambia la forma de construir conjuntos de datos para el aprendizaje supervisado. En estos casos, en lugar de tener conjuntos de datos etiquetados por una sola fuente (que se supone que es un experto que proporcionó el estándar de oro absoluto), se proporcionan bases de datos con múltiples anotadores. Sin embargo, la mayoría de los métodos de vanguardia dedicados al aprendizaje a partir de múltiples expertos suponen que el comportamiento del etiquetador es homogéneo en todo el espacio de características de entrada. Además, se imponen restricciones de independencia a los resultados de los anotadores. Este documento presenta una Red Neuronal Profunda Encadenada Regularizada para abordar tareas de clasificación a partir de múltiples anotadores. En esta tesis, desarrollamos dos estrategias con el objetivo de evitar las técnicas intrusivas que se utilizan habitualmente para diagnosticar motores de combustión interna (ICE). La primera consiste en una red neuronal de retardo temporal (TDNN), interpretada como un filtro de respuesta de pulso finito (FIR) para estimar la presión en el cilindro de un ICE de un solo cilindro a partir de las fluctuaciones de la velocidad angular del eje. Los experimentos se realizan sobre datos obtenidos de un ICE que opera en 12 estados diferentes cambiando la velocidad angular y la carga. El retardo de la TDNN se ajusta para obtener la mayor puntuación posible basada en la correlación. Nuestra metodología puede predecir la presión con un R2>0,9, evitando complicados pasos de preprocesamiento.
dc.descriptionMaestría
dc.descriptionMagíster en Ingeniería Eléctrica
dc.descriptionContent 1 Introduction 10 1.1 Problem statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 1.2 Justification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 1.3 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 1.3.1 General objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 1.3.2 Specific objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 2 TDNN-based Engine In-cylinder Pressure Estimation from Shaft Velocity Spectral Representation 18 2.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2.2.1 Time Delay Neural Network fundamentals . . . . . . . . . . . . . . . 19 2.2.2 Harmonic prediction performance based on Magnitude-Squared Coherence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.3 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.3.1 Engine Measurements, Data Acquisition, and Preprocessing . . . . . 22 2.3.2 Pressure signal estimation . . . . . . . . . . . . . . . . . . . . . . . . 26 2.4 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 2.5 Conclusions and future work . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 3 Master Thesis: Content 3 Regularized Chained Deep Neural Network Classifier for Multiple Annotators 37 3.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 3.2 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 3.2.1 Related work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 3.2.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 3.3 Experimental set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 3.3.1 Tested datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 3.3.2 RCDNN detailed architecture and training . . . . . . . . . . . . . . . 46 3.3.3 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 48 3.3.4 Introducing spammers and malicious annotators . . . . . . . . . . . . 55 3.3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 4 Final Remarks 58 4.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 4.1.1 TDNN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 4.1.2 RCDNN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 4.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 4.2.1 TDNN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 4.2.2 RCDNN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 4.3 Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
dc.format69 Páginas
dc.formatapplication/pdf
dc.formatapplication/pdf
dc.languageeng
dc.publisherUniversidad Tecnológica de Pereira
dc.publisherFacultad de Ingenierías
dc.publisherPereira
dc.publisherMaestría en Ingeniería Eléctrica
dc.relation[1] N. Lieven, P. E. Ambrosio, S. Burrow, and L. Clare, “Strategies for wireless intelligent sensing devices (wisds),” HUMS2007, Melbourne, AU, pp. 20–22, 2007.
dc.relation[2] V. M. Martinez, B. T. Martínez, P. C. O. González, and J. L. C. Velasco, “Diagnóstico de motores de combustión interna álternativos mediante el análisis de las oscilaciones del bloque,” Revista UIS Ingenierías, vol. 1, no. 2, pp. 11–15, 2002
dc.relation[3] A. S. Rosique, “Marco de referencia para la gestión de activos de alta capitalización: Definición de procesos de negocio y de técnicas avanzadas de soporte a la gestión,” Ph.D. dissertation, Universidad de Sevilla, 2017.
dc.relation4] H. Song and T. Jacobs, “The influence of soot radiation on no emission in practical biodiesel combustion,” Fuel, vol. 128, pp. 281–287, 2014
dc.relation[5] C. Bennett, J. Dunne, S. Trimby, and D. Richardson, “Engine cylinder pressure reconstruction using crank kinematics and recurrently-trained neural networks,” Mechanical systems and signal processing, vol. 85, pp. 126–145, 2017.
dc.relation[6] J. A. Grajales, H. F. Quintero, C. A. Romero, E. Henao, J. F. Lopez, and D. Torres, “Combustion pressure estimation method of a spark ignited combustion engine based on vibration signal processing,” Journal of Vibroengineering, vol. 18, no. 7, pp. 4237–4247, 2016.
dc.relation[7] G. Chiatti, O. Chiavola, E. Recco, A. Magno, E. Mancaruso, and B. M. Vaglieco, “Accelerometer measurement for mfb evaluation in multi-cylinder diesel engine,” Energy, vol. 133, pp. 843–850, 2017.
dc.relation[8] S. Narayan, S. Milojevic, and V. Gupta, “Combustion monitoring in engines using accelerometer signals,” Journal of Vibroengineering, vol. 21, no. 6, pp. 1552–1563, 2019.
dc.relation[9] G. Chiatti, O. Chiavola, E. Conti, and E. Recco, “Automotive turbocharger speed estimation via vibration analysis for combustion optimization,” Int. J. Mech. Eng. and Technology, vol. 8, no. 10, pp. 153–163, 2017.
dc.relation[10] K. Bizon, G. Continillo, E. Mancaruso, and B. M. Vaglieco, “Towards on-line prediction of the in-cylinder pressure in diesel engines from engine vibration using artificial neural networks,” SAE Technical Paper, Tech. Rep., 2013.
dc.relation[11] ——, “Reconstruction of in-cylinder pressure in a diesel engine from vibration signal using a rbf neural network model,” SAE Technical Paper, Tech. Rep., 2011.
dc.relation[12] L. Barelli, G. Bidini, C. Buratti, and R. Mariani, “Diagnosis of internal combustion engine through vibration and acoustic pressure non-intrusive measurements,” Applied Thermal Engineering, vol. 29, no. 8-9, pp. 1707–1713, 2009.
dc.relation[13] S. A. Ali and S. Saraswati, “Reconstruction of cylinder pressure using crankshaft speed fluctuations,” in 2015 International Conference on Industrial Instrumentation and Control (ICIC). IEEE, 2015, pp. 456–461.
dc.relation[14] Z. Boussaada, O. Curea, A. Remaci, H. Camblong, and N. Mrabet Bellaaj, “A nonlinear autoregressive exogenous (narx) neural network model for the prediction of the daily direct solar radiation,” Energies, vol. 11, no. 3, p. 620, 2018.
dc.relation[15] M. Längkvist, L. Karlsson, and A. Loutfi, “A review of unsupervised feature learning and deep learning for time-series modeling,” Pattern Recognition Letters, vol. 42, pp. 11–24, 2014.
dc.relation[16] J. E. Dayhoff and J. M. DeLeo, “Artificial neural networks: opening the black box,” Cancer: Interdisciplinary International Journal of the American Cancer Society, vol. 91, no. S8, pp. 1615–1635, 2001.
dc.relation[17] B. A. Toms, E. A. Barnes, and I. Ebert-Uphoff, “Physically interpretable neural networks for the geosciences: Applications to earth system variability,” Journal of Advances in Modeling Earth Systems, vol. 12, no. 9, p. e2019MS002002, 2020.
dc.relation[18] C. B. Azodi, J. Tang, and S.-H. Shiu, “Opening the black box: interpretable machine learning for geneticists,” Trends in genetics, vol. 36, no. 6, pp. 442–455, 2020.
dc.relation[19] M. T. Ribeiro, S. Singh, and C. Guestrin, “Model-agnostic interpretability of machine learning,” arXiv preprint arXiv:1606.05386, 2016.
dc.relation[20] C. Molnar, Interpretable machine learning. Lulu.com, 2020
dc.relation[21] J. Gil-Gonzalez, A. Orozco-Gutierrez, and A. Alvarez-Meza, “Learning from multiple inconsistent and dependent annotators to support classification tasks,” Neurocomputing, vol. 423, pp. 236–247, 2021.
dc.relation[22] R. Snow, B. O’Connor, D. Jurafsky, and A. Ng, “Cheap and fast—but is it good?: evaluating non-expert annotations for natural language tasks,” in EMNLP. ACL, 2008, pp. 254–263.
dc.relation[23] J. Zhang, X. Wu, and V. S. Sheng, “Learning from crowdsourced labeled data: a survey,” Artificial Intelligence Review, vol. 46, no. 4, pp. 543–576, 2016.
dc.relation[24] H.-E. Sung, C.-K. Chen, H. Xiao, and S.-D. Lin, “A classification model for diverse and noisy labelers,” in Pacific-Asia Conference on Knowledge Discovery and Data Mining. Springer, 2017, pp. 58–69.
dc.relation[25] D. Tao, J. Cheng, Z. Yu, K. Yue, and L. Wang, “Domain-weighted majority voting for crowdsourcing,” IEEE transactions on neural networks and learning systems, vol. 30, no. 1, pp. 163–174, 2018.
dc.relation[26] P. Ruiz, P. Morales-Álvarez, R. Molina, and A. K. Katsaggelos, “Learning from crowds with variational Gaussian processes,” Pattern Recognition, vol. 88, pp. 298–311, 2019.
dc.relation[27] S. Albarqouni, C. Baur, F. Achilles, V. Belagiannis, S. Demirci, and N. Navab, “Aggnet: deep learning from crowds for mitosis detection in breast cancer histology images,” IEEE transactions on medical imaging, vol. 35, no. 5, pp. 1313–1321, 2016.
dc.relation[28] F. Rodrigues and F. C. Pereira, “Deep learning from crowds,” in Thirty-Second AAAI Conference on Artificial Intelligence, 2018.
dc.relation[29] J. Gil-Gonzalez, A. Alvarez-Meza, and A. Orozco-Gutierrez, “Learning from multiple annotators using kernel alignment,” Pattern Recognition Letters, vol. 116, pp. 150–156, 2018.
dc.relation[30] M. Venanzi, J. Guiver, G. Kazai, P. Kohli, and M. Shokouhi, “Community-based Bayesian aggregation models for crowdsourcing,” in Proceedings of the 23rd international conference on World wide web. ACM, 2014, pp. 155–164.
dc.relation[31] W. Tang, M. Yin, and C.-J. Ho, “Leveraging peer communication to enhance crowdsourcing,” in The World Wide Web Conference. ACM, 2019, pp. 1794–1805
dc.relation[32] G. Vossoughi and S. Rezazadeh, “Optimization of the calibration for an internal combustion engine management system using multi-objective genetic algorithms,” in 2005 IEEE Congress on Evolutionary Computation, vol. 2. IEEE, 2005, pp. 1254–1261.
dc.relation[33] Y. G. Guezennec and P. Gyan, “A novel approach to real-time estimation of the individual cylinder combustion pressure for si engine control,” SAE Technical Paper, Tech. Rep., 1999.
dc.relation[34] H. Zhao, HCCI and CAI engines for the automotive industry. Elsevier, 2007.
dc.relation[35] S. Saxena and I. D. Bedoya, “Fundamental phenomena affecting low temperature combustion and hcci engines, high load limits and strategies for extending these limits,” Progress in Energy and Combustion Science, vol. 39, no. 5, pp. 457–488, 2013.
dc.relation[36] M. Wick, J. Bedei, D. Gordon, C. Wouters, B. Lehrheuer, E. Nuss, J. Andert, and C. R. Koch, “In-cycle control for stabilization of homogeneous charge compression ignition combustion using direct water injection,” Applied Energy, vol. 240, pp. 1061–1074, 2019.
dc.relation[37] H. Venu, V. D. Raju, S. Lingesan, and M. E. M. Soudagar, “Influence of al2o3nano additives in ternary fuel (diesel-biodiesel-ethanol) blends operated in a single cylinder diesel engine: Performance, combustion and emission characteristics,” Energy, vol. 215, p. 119091, 2020
dc.relation[38] A. O. Emiroğlu, “Effect of fuel injection pressure on the characteristics of single cylinder diesel engine powered by butanol-diesel blend,” Fuel, vol. 256, p. 115928, 2019.
dc.relation[39] B. Sudarmanta, A. Setiyawan, K. A. Bachtiar, D. Yuvenda, and J. da Silva, “Effects of advancing the pilot injection timing on the engine performance, combustion and emissions on diesel dual fuel engine at high load,” in IOP Conference Series: Materials Science and Engineering, vol. 588. IOP Publishing, 2019, p. 012020.
dc.relation[40] D. Gordon, C. Wouters, M. Wick, F. Xia, B. Lehrheuer, J. Andert, C. R. Koch, and S. Pischinger, “Development and experimental validation of a real-time capable field programmable gate array–based gas exchange model for negative valve overlap,” International Journal of Engine Research, vol. 21, no. 3, pp. 421–436, 2020.
dc.relation[41] X. Huang, L. Kang, M. Kassa, and C. Hall, “Cylinder specific pressure predictions for advanced dual fuel compression ignition engines utilizing a two-stage functional data analysis,” Journal of Dynamic Systems, Measurement, and Control, vol. 141, no. 5, 2019.
dc.relation[42] Y. Lee, S. Lee, K. Han, and K. Min, “Prediction of in-cylinder pressure for light-duty diesel engines,” SAE Technical Paper, Tech. Rep., 2019.
dc.relation43] J. Wu, A. Jacoby, D. Llamocca, and B. Sangeorzan, “A model for crank-angle-resolved engine cylinder pressure estimation,” SAE Technical Paper, Tech. Rep., 2018.
dc.relation[44] A. Evlampiev, L. Somers, R. Baert, and L. De Goey, “On the impact of the ideal gas assumption to high-pressure combustion phenomena in engines,” Combustion science and technology, vol. 180, no. 2, pp. 371–390, 2007.
dc.relation[45] Z. Yue, R. Hessel, and R. D. Reitz, “Investigation of real gas effects on combustion and emissions in internal combustion engines and implications for development of chemical kinetics mechanisms,” International Journal of Engine Research, vol. 19, no. 3, pp. 269–281, 2018.
dc.relation[46] Y. Yeliana, C. Cooney, J. Worm, D. Michalek, and J. Naber, “Estimation of double-wiebe function parameters using least square method for burn durations of ethanol-gasoline blends in spark ignition engine over variable compression ratios and egr levels,” Applied thermal engineering, vol. 31, no. 14-15, pp. 2213–2220, 2011.
dc.relation[47] F. Liu, G. A. Amaratunga, N. Collings, and A. Soliman, “An experimental study on engine dynamics model based in-cylinder pressure estimation,” SAE Technical Paper, Tech. Rep., 2012.
dc.relation[48] V. C. Mariani, S. H. Och, L. dos Santos Coelho, and E. Domingues, “Pressure prediction of a spark ignition single cylinder engine using optimized extreme learning machine models,” Applied energy, vol. 249, pp. 204–221, 2019.
dc.relation[49] S. Polat, Ö. Özdilli, and H. Cizmeci, “An estimation of incylinder pressure based on lambda and engine speed in hcci engine using artificial neural networks,” Fresenius Environmental Bulletin, vol. 28, no. 4A, pp. 3568–3576, 2019.
dc.relation[50] S. Trimby, J. F. Dunne, C. Bennett, and D. Richardson, “Unified approach to engine cylinder pressure reconstruction using time-delay neural networks with crank kinematics or block vibration measurements,” International Journal of Engine Research, vol. 18, no. 3, pp. 256–272, 2017.
dc.rightsManifiesto (Manifestamos) en este documento la voluntad de autorizar a la Biblioteca Jorge Roa Martínez de la Universidad Tecnológica de Pereira la publicación en el Repositorio institucional (http://biblioteca.utp.edu.co), la versión electrónica de la OBRA titulada: ________________________________________________________________________________________________ ________________________________________________________________________________________________ ________________________________________________________________________________________________ La Universidad Tecnológica de Pereira, entidad académica sin ánimo de lucro, queda por lo tanto facultada para ejercer plenamente la autorización anteriormente descrita en su actividad ordinaria de investigación, docencia y publicación. La autorización otorgada se ajusta a lo que establece la Ley 23 de 1982. Con todo, en mi (nuestra) condición de autor (es) me (nos) reservo (reservamos) los derechos morales de la OBRA antes citada con arreglo al artículo 30 de
dc.rightsinfo:eu-repo/semantics/openAccess
dc.rightshttp://purl.org/coar/access_right/c_abf2
dc.rightsAtribución-NoComercial-SinDerivadas 4.0 Internacional (CC BY-NC-ND 4.0)
dc.rightshttps://creativecommons.org/licenses/by-nc-nd/4.0/
dc.subject620 - Ingeniería y operaciones afines::629 - Otras ramas de la ingeniería
dc.subjectSistemas de control de supervisión
dc.subjectRedes neuronales (Computadores)
dc.subjectInteligencia artificial
dc.subjectAprendizaje de máquina
dc.subjectMotor de combustión
dc.subjectMúltiples anotadores
dc.titleDevelopment of a methodology for the diagnosis of internal combustion engines using non-invasive measurements based on the use of interpretable neural networks applicable to databases with multiple annotators
dc.typeTrabajo de grado - Maestría
dc.typehttp://purl.org/coar/resource_type/c_bdcc
dc.typehttp://purl.org/coar/version/c_ab4af688f83e57aa
dc.typeText
dc.typeinfo:eu-repo/semantics/acceptedVersion


Este ítem pertenece a la siguiente institución