Internal combustion engine analysis using EngineEnvelope (EE) L. Bregant 1, M. Spagnol 1, G. Caizzi 1 1 Dipartimento di Ingegneria e Architettura, Università degli Studi di Trieste Via Valerio 10, 34100 Trieste, Italy e-mail: bregant@units.it Abstract Many common industrial activities, like energy production, people and goods transportation, heavily rely on internal combustion engines operations. The capabilities of reducing the total cost of operation and maximizing the uptime of these machines are a very important topic for the their owner and much energy has to be put on the purpose. In many cases, scheduled maintenance programs are adopted. After a defined amount of time, decided by the engine producer, parts are overhauled and updated regardless of the real usage level and damage. Instead, it is clear that being able to substitute only the damaged parts, at due time, will reduce the engine ownership and maintenance costs. While on the market several maintenance tools dedicated to the on-line monitoring of the thermodynamic aspects (mixture characteristics, exhausts analysis, miss-firing) are present, none are available for the evaluation of the kinematics links problems. The paper propose a novel technique, named EngineEnvelope (EE), focused on the online monitoring of the mechanical potential issues like bolt looseness, pistons clearance and bearings damages. The paper describe the monitoring function, its capabilities of detecting damages on real data acquired on a two strokes, single cylinder engine. 1 Introduction The internal combustion engines are very common in many industrial sectors and applications: from the power generation plants to the goods moving machines from people transport fleets to energy source form operating machines. Due to their widespread availability, a lot of attention is devoted to the maintenance of those machines. Programs of scheduled activities, in which different components are overhauled, regenerated or changed are foreseen from the engine producers, in order to maintain the performances and the efficiency at the highest level. Those program, tough, do not contain indication about the kind of load these engine are subjected to during their operation. The time is the only variable by which those stops are programmed. It is evident that, different loading conditions will act differently on the components of the engine and the maintenance should be scheduled as a function of those loads. Failing to do this, the substitution of the overhauled component is a total waste of resources because it is still operational. That is the reason why a condition based monitoring is preferable in those machines in which the operating conditions are variables as the ones typical of the IC engines. Furthermore, while there are different sensors for the real-time evaluation of the thermodynamic cycle (e.g. the lambda sensor to evaluate the contents of oxygen in the exhausts) seldom accelerometers are used, on-line, for the evaluation of the mechanical dynamic of the system. Knocking, piston slap, misfiring, bearing defects and bolt looseness instead could be identified and localized with the appropriate measurements and data analysis techniques. The present paper focuses on a newly developed technique for the mechanical monitoring of IC engines called EngineEnvelope (EE). This technique is based on the analysis of accelerometers data acquired in some especially selected points of the engine, and it has been optimized to cope with the inherent RPM variations of an engine. It provides a statistical indication on the engine vibration level as function of the crankshaft angle and as such pinpointing with an high degree of confidence where the malfunction could 2787
2788 PROCEEDINGS OF ISMA2014 INCLUDING USD2014 be located. It will be shown, on a two strokes single cylinder engine, how accurate the result of EE can be. Two different loading conditions were tested: dragged, to simulate the End Of Line testing; autonomously operated, to reproduce the ordinary operating conditions. Different faults will be inserted in the engine and the variation of the EE function highlighted. 2 Internal Combustion Engine Kinemetics 2.1 Review Internal combustion engines have a know characteristics kinematics: the pistons move in a linear fashion within the cylinders, while the crankshaft rotates around the basement bearing. The piston rod, connecting the two, allows a double translation s inversion for the piston for each complete crankshaft revolution. This, combined with the firing orders in the multi-cylinders engine, yields for the appearance of multiple orders in the vibration signals measured on the IC engines. On top of these orders, many different ones will appear due to any malfunctioning, like the knocking (local combustion of the air-fuel mixture) the piston slap (vibration of the piston, within the cylinder, around the connecting road pin) the misfire (lack of one or multiple combustions), the valve timing (wrong closing and opening of the valve for intake and discharge of the combustion fluids), the bearing faults (either needle bearing or hydrostatic bearing), the block vibration (on the supports) and the engine head s bolts looseness (reducing the compression values of the air fuel mixture). Several authors in the past dealt with these kind of problem and signals processing, [1]-[7]. What appears evident, thanks to the cyclic functioning of the machines, is that each of the mentioned faults, have a specific signature and appearance in the 360 (2-strokes) or 720 degree (4- strokes) plots and these peculiarities are what EE is trying to highlight. Figure 1 shows a simple engine schematic, a valve timing diagram and the first order evolution as function of the engine RPM. The capabilities of evaluating the health of a machine and its components relays on the accurate measurements of two quantities, the instantaneous speed and its vibration in appropriate points. Bringing this to the industrial environment, the definition of simplified procedures and automatic tools is required. EE consists of different steps, namely the RPM acquisition, the sensors placement, the EE function evaluation and the defect diagnostics. In the following all three will be briefly described. Figure 1: IC engine basics
MONITORING AND DIAGNOSTICS OF ROTATING MACHINERY 2789 3 Engine Envelope 3.1 RPM acquisition and AutoTacho RPM sensors are commonly available devices, any 1x or more pulses per revolution will do. In order to minimize the number and type of the requested sensors, the instantaneous speed of the engine is automatically computed from the acceleration signal with the AutoTacho procedure described in the following. The technique, is less accurate than the pulse based measurement, but for contained speed variation rates, it provides accurate enough information. Furthermore, it is worthwhile remembering that during run-up and coast-down measurement, the duration of the ramp influences the accuracy of the system dynamics estimates [8]. The AutoTacho procedure requires the following steps: spectrogram evaluation maximum order identification and tracking maximum order correction with AdjacentOrders RPM evaluation The procedure can be fully automated. The parameters to insert are the ones for the spectrogram evaluation and the number of AdjacentOrders used into the optimization phase. An high number of frequency lines on which the FFT is computed (about 2^16) will yield to a high frequency resolution, while an high overlap (about 95%) assures a good tracking of the speed variations, two to four AdjacentOrders will lead to accurate results in most of the case. With the suggested parameters the computational burden is significant, but the quality of the estimation greatly improves. It is worthwhile remembering that the RPM evaluation is performed once for each run, and the quality of the subsequent evaluation depends from the quality of the instantaneous velocity estimation. The correction of the tracked order with the Adjacent ones is performed searching the maximum value of the order between the tracked order, multiplied for the appropriate constant, and the measured corresponding order. (e.g.: IC engines have strong order II, this can be tracked as the maximum order and the RPM is evaluated. The maximum order is multiplied by a factor 2, compared with the computed order IV. The maximum value between the two is kept and RPM values calculated. The maximum order is multiplied by a factor 3, compared with the computed order VI and so on). The differently obtained RPM estimates are than averaged to obtain the actual RPM profile. The results can be seen in figure 2. (a) is the calculated spectrogram, with highlighted AdjacentOrders, (b) is the acquired vibration signal and (c) is the obtained calculated RPM profile with the maximum order alone (red) and corrected with two extra AdjacentOrders (green). Figure 2: Autotacho output. (a) Spectrogram, (b) Vibration Signal, (c) RPM profile
2790 PROCEEDINGS OF ISMA2014 INCLUDING USD2014 Figure 3: Engine Segmentation and OrderSurface sensors placement map 3.2 Sensor Location Once the RPM profile is estimated, it is required to optimize the sensors positions in order to acquire the response of the system with the highest dynamic. In general, a trial and error procedure could be adopted considering the limitations of the chosen sensors. The high temperature is the most critical for the survivability of the equipment, while the mounting technique is the second most challenging problem. In fact, most of the IC engine block are non magnetic alloys. Stud mounting, due to the expectable acceleration levels would be the preferred choice, but need to be designed beforehand. A part from the practical considerations, the position of the sensor will limit or enhance its capabilities of detecting damages, and a dedicated procedure has been devised, based on evaluation of the OrderSurface. This surface is obtained evaluating the amplitude of the order of interest on a measurement points grid on the engine block. This surface can be evaluated for the different orders (integer or non integers) and different directions as function of the most seeked damage. (e.g. looking for the piston slap occurrence, will lead to a measurement point close to the engine head, in a direction orthogonal to the cylinder and crankshaft axis). For the single engine case, most of the information can be gathered from the lower cylinder area as figure 3. This shows the engine segmentation and the OrderSurface from order 1 in the Z direction. It appears evident that the best location to measure the first order is located in the lower part of the crankcase, in middle plane. A refined mesh would not be needed, but on multi cylinders engine, the mapping ought to be performed on several different planes. 3.3 Engine Envelope With the appropriate RPM profile and the vibration measurements locations, the EE function can be evaluated for each point and directions. Normally a limited number of points (1-2) will be sufficient to characterize such an engine. For the multi cylinders cases a higher number is requested to encompass the whole system dynamics (at least a sensor each two cylinders). The calculation of the EE function requires the following steps: angular resampling of the vibration signal, to eliminate the RPM variation, time synchronous averaging of the resampled signal, to reduce the random errors upper and lower envelope estimation average estimation between the envelopes normalization of the signal with respect to the signal average Figure 4 shows the EE function calculation development. Figure 4a represent the original data, Fig. 4b the normalized version, fig. 4c the EE function with the confidence interval of +/- σ. The obtained EngineEnvelope function is plotted against the crankshaft angle with the confidence interval depending from the number of averages on which the Time Synchronous Averaging (TSA) has been computed.
MONITORING AND DIAGNOSTICS OF ROTATING MACHINERY 2791 Figure 4c show the final EE function for the single cylinder case. Obviously the peak position, corresponding to the air-fuel mixture combustion, is phased either with a key-phasor, if available, or shifted on the graph according to the phase plot of the engine. 4 Measurements Results The damage detection capabilities of EE have been thoroughly tested modifying the working conditions and the mounting characteristics of the chosen single cylinder, two stokes motorcycle engine. Several damages have been simulated removing piston s rings ( n 1 closest to the piston head, n 2 closest to the piston pin), loosening head s bolts, mechanically deforming the piston pin bearing, both in thermal and dragged condition. In each cases, differences could be highlighted in the EE function, and the entity of the damage could be evaluated analyzing the relative positions of the confidence intervals. Figure 5 shows how the piston slap changes as function of the piston ring presence. The EE function in blue is the correct engine case, while the red represents the modified response. The measurement is in orthogonal direction with respect to cylinder axis (radial direction). The main information are: valve opening at 300-330, Top Dead Center (TDC) 180. Just Before of the TDC, the acceleration increases due to the piston slap phenomena: the rod is near the maximum extension so the clearance between piston and cylinder is amplified. This effect is more evident without the piston ring 2. At the inlet port s opening, just before 300, the vibration signal is reduced maybe due to a kind of air cushioning made from the absence of the piston ring. Figure 6 exhibits the EE function for the damage connecting rod-piston needle bearing: in this case a strong increment in vibration level can be seen between 200-240, while at 300 the level is reduced. Figure 7 shows the difference in the EE function when one of the piston head bolts is loosened from 12Nm (original value) to 5Nm. The acceleration value increases when the piston reaches the extremes of the stroke due to the inversion of its running direction. The head s movement generates many impacts and a leakage of exhaust gasses is present. Figure 4: EE function development (a) original signal, (b) normalized version, (c) EE function with +/- σ
2792 PROCEEDINGS OF ISMA2014 INCLUDING USD2014 Figure 5: EE function for piston ring removal, blue original, red modified, (a) no piston ring1, (b) no piston ring2 - dragged engine Figure 6: EE function for bearing deformation, blue original, red damaged,- dragged engine Figure 7: EE function for bolt loosening, blue original, red damaged,- normal thermal operation engine
MONITORING AND DIAGNOSTICS OF ROTATING MACHINERY 2793 5 Conclusions In this paper a new statistical method for acceleration data analysis is presented. The experimental results confirm the capability of EE to identify the defect presence. A baseline measurement is requested in order to have a condition as new of the motor. So this method is suitable for monitoring the entire life of an engine, from commissioning to its end of life. References [1] R. B. Randall, Vibration-based Condition Monitoring, West Sussex, Whiley, 2011. [2] S. H. Cho, S. T. Ahn e Y. H. Kim, A simple model to estimate the impact force induced by Piston Slap, Journal of Sound and Vibration, Vol 255, n 2, (2002), pp.229-242. [3] Z. Geng and J. Chen, Investigation into Piston-Slap-Induced vibration for engine condition simulation and monitoring, Journal of Sound and Vibration, Vol 282, n 3-5, (2005), pp.735-751. [4] M. El. Badaoui, J. Daniere, F. Guillet and C. Serviere, Separation of combustion noise and pistonslap in diesel engine Part I: Separation of combustion noise and piston-slap in diesel engine by cyclic Wiener filtering, Mechanical Systems and Signal Processing, Vol.19, n 6, (2005), pp.1209-1217. [5] J. M. Lujàn, V. Bermùdez, C. Guardiola and A.Abbad, A methodology for combustion detection in diesel engines through in-cylinder pressure derivative signal, Mechanical Systems and Signal Processing, Vol.24, n 7, (2010), pp. 2261-2275. [6] Y.V.V. Satyanarayana Murthy, Combustion Analysis and Knock Detection in Single Cylinder DI- Diesel Engine using Vibration Signature Analysis, Internation Journal of Engineering Science and Technology, Vol.3, n 1, (2011), pp 10-16. [7] J. Antoni, J. Daniere and F. Guillet, Effective Vibration Analysis of IC engines using Cyclostationarity. Part II- New Results on the reconstruction of the Cylinder Pressure, Journal of Sound and Vibration, Vol.257, n 5, (2002) pp.839-856.
2794 PROCEEDINGS OF ISMA2014 INCLUDING USD2014