The paper presents a possible method to diagnose a mechanical fault of an automotive system. Starting from the point of view that every fault of a mechanical system should introduce an abnormal component within the signal that describes the time history of a mechanical parameter we tried to find a way to reveal it.We were performing some tests involving a military vehicle with respect to the performances of its braking system. The tests were aiming at identifying a way to bring up-to-date the old weapon system from the braking systems point of view. During these tests we observed some anomalies concerning the pressure evolution within the braking cylinders of the vehicle. Some unusual but also systematic noises occurred. As a main issue at this point, the source of the noise should have been identified and filtered if necessary. We had to decide whether the noisy component of the signal is just a noise that should be removed by filtering the signal or it is a physical component of the mechanical parameter itself (not noise but a useful information).These procedures take time and they also request accurate knowledge as well as fine expertise in automotive testing. Since our Dept. has a long and rich practice in this respect, we assumed to processing data and give them a thorough interpretation. So, the first thing we did was to perform a frequency analysis, using classical methods. Usually, a simple frequency analysis cant provide information about a time variation of the frequency spectra due to the Fourier Transforms behavior, since it freezes the signal in time. A much more accurate analysis is the time-frequency analysis. However, observing both the amplitude and power spectra can lead to a useful conclusion. We concluded that the noise we met within the signal is due to the brake drums loss of circular shape (they turned into an oval, the process being known as ovalization). Hence, we cant talk about a noise as it is usually defined.

The simulation procedure has always been considered as a giant leap forward, especially in the field of basic designing of a product. There is nothing new underneath the basic concept, but the scientific and technical progress always brings up new techniques that improve simulation in its whole. When we talk about a vehicle, especially about a military one, we consider that it is much cheaper to simulate a process involving the weapon system than performing countless tests that are rather expensive. In this respect, we tried to develop a simulation mathematical model, check its accuracy with a set of extensive tests, prove it reliability and further extrapolate the behavior of the simulated model to a larger number of military vehicles of the same kind. It could help in various fields, such as diagnose (by comparing the simulated results with the real ones got from a faulty vehicle) or automatically regulating some functions (an intelligent vehicle, having an implemented, simulated model, that is able to feel the status of a subsystem in real time and regulate its behavior, accordingly). Hence, the paper presents a model that simulates the longitudinal dynamics of a tracked vehicle. It has been issued using Simulink module of Matlab programming environment. It aims at pointing out the performances of the vehicle. The models interface is friendly and its structure is modular. The main modules of the model are the engine, the torque converter, the transmission and the track. The engine and the torque converter are modeled using the experimental maps obtained by the tests that have been previously developed by the manufacturer. The main principle of the equations that describe the system is to set a balance among the forces (both active and resistive) that load the vehicle. The inputs of the model are the technical and dimensional features, provided by the manufacturer or experimentally determined. The output of the model is a dynamic behavior. Comparing the results with the experimental data eventually validates or invalidates the model. But the results were excellent, so the model was validated. Also, the results proved that the developed model is able to predict the performances of the take-off stage of the tracked vehicle.

The paper deals with a new approach in data analysis of a measured mechanical parameter. The classic approach is mainly based on the deterministic statistics that cant cover the whole field of a complete analysis. The stochastic approach, to be used in this paper, offers far more information about the mechanical parameter and can take into account the non-linearity of the signal, eventually, the mechanical parameter itself. Starting from the point of view that, in real life, there is no steady evolution of any parameter, we decide to take into account the importance of the non-linear components of any signal. After e thorough investigation, we hope we could make the difference between the noise, as non-linear components of the measured parameter, and the useful non-linear components (e.g. important shocks, typically met within a vehicles transmission). Using the stochastic modeling procedures, we aimed at issuing comprehensive, accurate and valuable dynamic models of the phenomenon. These models cam be used in a large variety of situations, from describing the process, to evaluating the health of a mechanical system and to controlling a real-time process based on the pre-set models (previously drawing a map of the systems normal behavior and permanently assessing the deviation from it and acting accordingly). The data were measured within the transmission system of a military vehicle. Specifically, we have gathered information about torque and angular speed of different shafts of the driveline. As everybody knows, the power flows within any vehicles transmission in transient modes mainly and it is accompanied by plenty of noise. It is rather challenging to separate (filter) the useful signal form the noise but, on the other hand, it is the only way to achieve useful data. Therefore, a spectral analysis is a must, but not the conventional one, which has its drawbacks, but a multi-spectral one, which is able to insulate the noise. Moreover, starting from the analysis developed with this method, mathematical models, both in discrete and continuos time can be achieved. It is easy to notice that the models that we have achieved are featured by a very good accuracy. We could push the data processing even further, getting generalized models that provide the needs we have mentioned before, with respect to the mapping of a normal (averaged) behavior of a system, to be used in controlling procedures.

No abstract

This paper is the result of a research program which focused on the statistical dynamics of vehicles. Most of the inputs of man-machine-field system have a random variation, so a systemic and statistical analysis of vehicle dynamics is obvious. In our study, data were obtained by measuring the dynamic parameters of vehicles and engines. Testing program aimed to capture a large range of operating regimes. To analyze the data the authors have used neural networks. There was adopted a NNARX (Neural Network Auto-Regressive with eXogene inputs) model with 4 inputs, 5 hidden units and 1 output. It can be concluded that the development of mathematical modeling using non-linear neural network can ensure the desired accuracy, conveniently is obtained by increasing the number of neurons in the hidden laws.