روشی برای شناسایی ترک ها در میل لنگ بزرگ

Diesel engines used in power plants and marine propulsion are especially sensitive to outage events. Any advance in the early detection of failure will increase the reliability of the electricity supply and improve its productivity by reducing costly power outages. Fault detection and diagnosis is important technology in condition-based maintenance for diesel engines. This article presents a classifier based on neural networks for identifying failure risk level in crankshafts, the engine component of greatest cost concern. The authors have developed a finite element model for crack growth that fits well with fracture appearance and produces the evolution of crankshaft stiffness with crack depth. A lumped system model of the engine uses this evolution as input, giving the instantaneous speed at the engine flywheel as a function of crack depth. All the results shown in the paper come from outputs of the simulation models which have been built from real engine data. Measurements of the instantaneous flywheel speed were not available due to the crankshaft failure. All data are extracted from this speed and are then classified using a Radial Basis Function neural network.

Diesel engines are commonly used in power plants all around the world, particularly for on-site power generation in special locations such as islands, which are not connected to a mainland electrical grid. This kind of power plant is especially sensitive to outage events. Any advance in the early detection of failure will increase the reliability of the electricity supply and will improve its productivity by reducing costly power outages. This is of particular interest in order to detect problems related to the engine crankshaft. In the case of crankshaft failure, repair costs include not only that of the crankshaft itself, but the cost of other parts of the engine that can be affected by crankshaft failure, such as connecting rods, pistons and cylinders, must be added. In addition, the length of time required for repairs has to be taken into account, mainly because of the crankshaft location inside the engine. This greatly increases the total repair cost. Several reliability, availability and maintainability (RAM) studies of diesel generators have been conducted and in some of them [1], statistics on availability, failure cause, mean time between forced outages and so on have been shown. In relation to diesel engines, Arinc Research Corporation conducted a study for the US Army Engineering and Housing Support Centre (EHSC) [2] that showed results from diesel engines up to 2 MW. This study included a detailed classification of the parts involved in the failure of this power range and it revealed that, even though the failures per year related to engine crankshaft were low (0.02), it resulted in a higher mean time to perform corrective maintenance (MTTCM) for outages due to crankshaft, see Table 1[2]. (MTBF=mean time between failures; MTBCM=mean time between corrective maintenance; MTTCM=mean time to perform corrective maintenance.) Similar conclusions are shown in Ref. [3]. The most common cause of crankshaft failure is fatigue. In order for fatigue to take place, a cyclic tensile stress and crack initiation site are necessary. The crankshafts of diesel engines of power plants run with harmonic torsion combined with cyclic bending stress due to the radial loads of the combustion chamber pressure transmitted from the pistons and connecting rods, to which inertia loads have to be added. Although crankshafts are generally designed with a high safety margin in order not to exceed the fatigue strength of the material, high cyclic loading and local stress concentration could lead to the growth of cracks even when fatigue strength is not exceeded by average values. Pandey [4] analyses failures in the crankshafts of 35 hp two-cylinder engines used in tractors, where the fracture plane was located between the main bearing and the journal. The crack was initiated at the crankpin web region in a plane about 45° in respect to the rotational axis, showing typical fatigue failure with beach marks. The stress related to the fatigue initiation was estimated at 175 MPa, significantly below the tensile stress of the nodular cast iron of these crankshafts which is close to 680 MPa. Taylor et al. [5] developed two fatigue experiments in a crankshaft of a four-cylinder engine made of spheroidal graphite cast iron, with a tensile strength of 440 MPa: one torsional and the other flexural. The crankshafts underwent torsional and flexural cyclic loading until failure and in both types of tests the same fracture angle of 45° in respect to the rotational axis was observed. Yu and Xu [6] investigated the fracture of the web between the 2nd journal and the 2nd crankpin of the crankshaft of a four-cylinder diesel engine of a truck plant. The failure occurred after 200 h in service and the fracture plane was about 45° inclined with respect to the shaft axis. The macroscopic view of the fracture surface indicated stable crack growth regions with beach marks in the middle. Baumik et al. [7] studied the fracture of the crankshaft of a four-cylinder aircraft engine made of case-hardened SAE 4340 grade forged steel. The fracture had taken place along the webs at the No. 2 and No. 3 journals after 1460 h in service and 262 h since the last overhaul. In both cases, the fracture was produced along the web radius, and transverse to the axis of the crankshaft. In journal 3 the fatigue crack had propagated to about 80% of the web cross-section before giving rise to the final overload fracture. In these cases it was possible to discover the origin of the fracture by tracing back the beach marks, which was found to be at the web radius region. Other investigations related to crankshaft failures gave similar results [8]: crankshaft failure is due to fatigue that is initiated by cracks located at the web fillet radius and progress to the journal inner radius leading to the final overload fracture. In the case of cracks in rotating structures, one of the approaches used to identify them is based on the fact that the presence of a crack reduces the stiffness of the structure, hence reducing the natural frequencies of the original healthy shaft without cracks. The change in modal properties, natural frequencies and mode shapes, may be useful for the detection of a crack as well as its depth and location [9], [10], [11], [12], [13], [14], [15] and [16]. Currently, intelligent optimisation techniques have been included in fault recognition methodologies applied to engines, especially artificial neural networks (ANN) with very interesting results [17], [18], [19], [20] and [21]. ANN techniques can handle incomplete data to deal with nonlinear problems and, once trained, can perform predictions and generalisations with important time savings. Among the different existing architectures and methodologies, Radial Basis Function (RBF) neural networks correspond to a very useful kind of ANN for classifying applications and function approaches [22] and [23]. The aim of this paper is to show a methodology for the identification of cracks and their depth in crankshafts, which will allow improvement in the predictive maintenance strategy of diesel power plants. The methodology is based on the development of a dynamic model of the crankshaft coupled to a 3D FEM model of the crack growth, applying a RBF neural network for classification purposes. Although FEM dynamic models lead to more precise solutions than the models built with a lumped-parameter system [24], good results, as well as their low computational cost based on inertia–spring–damper elements for each degree of freedom, have resulted in their being chosen for the study. The classification method is based only on lumped model results. finite element model allows calculating crank stiffness and crankshaft stress for each crack growth step. Lumped system model takes these data from FE results and then it is feasible to establish the relationship between crack growth and maximum angular flywheel speed oscillation (MAFSO). The work shown in the paper analyses thirteen crack growth step which cover from zero crack depth till the maximum section area before failure. This area is close to 43% of the crack growth area at 30° of crack plane angle (respect crankshaft axis). This limit was estimated from inspection of progression marks at the failure surface. The detection tool, which is the core of the diagnosis system for crack identification, is based on the comparison of MAFSO patterns obtained from the lumped system model for each engine load versus instantaneous measured MAFSO at the flywheel. This method is currently working properly in a real engine. The method is able to predict the initiation of a crack and its depth through the measurement of the instantaneous engine speed at the flywheel, and it has been developed from original data of the diesel engine generator described below. Although part of the information related to the diagnosis system has been exposed in the paper, due to commercial constraints some of the development cannot be published.

The methodology presented here for crack growth modelling based on a FE model shows very good concordance with failure surface appearance, which validates the failure identification procedure. Both crack growth planes analysed, those of 30° and 45°, show quite a similar relationship between journal stiffness and the increment in the maximum amplitude of the flywheel speed oscillation (MAFSO). This result increases the generality of the method because, as has been shown in the introduction, all the cracks develop according to a growth angle of between 30° and 45°. The method has shown its capability for crankshaft crack identification and depth evaluation for any crack location. The completely developed procedure allows the classification of the crack risk of a power plant engine exclusively due to crankshaft growth area into three levels of damage. The high degree of effectiveness of the results presented demonstrates that the designed RBF neural network constitutes a very promising tool for predicting engine cracks and for maintenance operations necessary in order to avoid engine cracks. This tool can be integrated in a system on-line identifying risks at real-time. This study will be completed with further work in two areas: first, incorporating combustion malfunctions into this analysis could extend the diagnosis to combustion fault recognition, as well as crankshaft failure; second, identifying the crankshaft journal that suffers from crack growth.