We present a model-based approach for the instant detection of faults on the DC side of photovoltaic (PV) systems. The algorithm does not identify the faults itself, but estimates the nominal PV system behavior, i.e. system parameters, using simple PV and line models. Sudden deviations from the expected model behavior serve as an indicator for the ignition of a fault. To ensure that the PV model parameters can be estimated, an identifiability analysis has to be performed. The performance of the algorithm is demonstrated exemplarily by the detection of serial electric arcs in PV systems. Measurement results show that all series arc faults are successfully detected. There are no false detections due to maximum power point tracking (MPPT) operations or environmental influences like shading, changes in solar irradiation, etc. The main advantages of the presented method are less computational effort, resulting in very fast detection times, and its flexible integration into existing systems.

In course of the continuing trend towards lightweight construction of fast positioning systems, cable feeders, sensor connections, power supply, etc., may have noticeable influence on the system behavior.
To study the effects of such additional masses a multi-body model of a positioning drive, serving as a reference model, is built and compared with measurements. Since it turns out that this multi-body model is unsuitable for the development of systems identification methods we strive for finding a proper, concentrated parametric mean value model that takes into account the variable mass. Finally, this model is used to test and evaluate standard methods for parameter identification based on Poisson moment functions.

Vehicle dynamics control (VDC) for motorcycles had a fast growth during the last 10 years. The available technologies comprise curve-safe ABS and traction control (TC) systems, anti-wheelie control, right up to comprehensive motorcycle stability systems including even more control functions. VDC systems rely on real-time information about the current motorcycle dynamic state. Thus motorcycles are equipped with additional sensor units, namely MEMS inertial measurement devices, capable of gathering accelerations and angular rates. The application of model-based estimation theory enables the determination of the necessary information about the in-plane and out-of-plane motion, e.g. the motorcycle lean angle. Since VDC systems include safety critical control functions, the validation within simulations including sensor characteristics is mandatory. The MEMS accelerometer and gyroscope features include low-cost and small footprint, however there are considerable stochastic sensor errors to cope with. In this study the characteristic of different MEMS sensors and their noise models are investigated. The sensor noise terms are identified by analyzing measurement data using the Allan variance method. Different sensors are compared and the stochastic noise coefficients are quantified. The sensor noises are modeled with according random processes defined by linear time-invariant systems and white-noise inputs. As a result, the obtained stochastic sensor models can be used for model-based estimation and control algorithm design, as well as verification within simulation environments.

In this study we focus on systematic disturbances caused by the motorcycle pitch dynamic when measuring longitudinal acceleration on motorcycles using low-cost acceleration sensors. Major systematic influences in the sensor measurement like gravitational acceleration, suspension dynamics and the road slope are addressed. During acceleration phases the motorcycle pitch angle changes according to the suspension setting. As a result the longitudinal sensing axis of the accelerometer includes parts of the gravitational acceleration and lags parts of the longitudinal acceleration. Gravitational acceleration has also significant influence on inclined roads. To obtain correct values of the effective longitudinal acceleration, the disturbances in the measured signal are analyzed and in further consequence compensated. For this purpose a linearized in-plane-dynamics model of the motorcycle is derived from a comprehensive multibody simulation. The mathematical description of the motorcycle behavior includes the systematic influences and serves as a basis for the model-based compensation. A state observer for pitch angle estimation and road slope reconstruction is designed. As a result the measured acceleration can be corrected with the estimated quantities. In the course of ongoing development of motorcycle dynamics control and advancement of drivetrains, e.g. hybridization, independent and economically measureable reference signals are required to achieve desired vehicle behavior with closed-loop control. Acceleration measurement on two-wheelers can be achieved at low hardware cost and is a promising quantity for innovative control designs.

We study optimal input design and bias-compensating parameter estimation methods for continuous-time models applied on a mechanical laboratory experiment. Within this task we compare two online estimation methods that are based on Poisson moment functions with focus on quantized system outputs due to an angular encoder: The standard recursive least-squares (RLS) approach and a bias-compensating recursive least-squares (BCRLS) approach. The rationale is to achieve acceptable estimation results in the presence of white noise, caused by low-budget encoders with low resolution. The input design and parameter estimation approaches are assessed and compared, experimentally, resorting to measurements taken from a laboratory cart system.

In this paper we present a free tool for semi-automated matching of virtual and real prototypes in a wide range of industrial applications. Within the introduction we would like to explain the motivation behind the development of our tool. After a short description of the basic principles of input design for dynamical systems, we are focusing on the individual steps of the tool-chain, implemented in our software.

In this study, we present an adaptive self-tuning controller (STC) design for small embedded systems. The new free tool-chain for model based control design is based, among other software, on the open simulator Scilab-XCos. After a very short introduction of model based design terms, this article focuses on the code generator and the other programs of the tool-chain. The design concept is demonstrated by the non-trivial adaptive self-tuning control (STC) of the cart system in simulation and on a real laboratory experiment.

Several papers have been published in the past on the issue of decomposing a nonlinear system into subsystems for more efficient time integration. In this paper each body of a multi body system is considered as one subsystem. The subsystems (the bodies) are interacting via connection forces. The sources of such connection forces are constraints or directly applied forces. This contribution is restricted to constraint forces only. During a step which is named “body iteration”, those forces are considered as constant and the state of the system is computed for each body separately. This can be massively parallelized which can be an efficiency advantage in case of computational costly problems like the ones occurring in parameter estimation. During an “constraint update step” the constraints are evaluated based on the body’s current state. If the error is not small enough the interface forces are updated and the inner loop is executed once again until the error of the constraints is negligible. It turns out, that the constraints can be updated separately as well, which can be used again for parallel computing. In the paper, the theory will be outlined and implemented using an N body pendulum. Finally, the advantages and disadvantages of this approach are critically discussed.

Flexible multibody simulation, subject to holonomic constraints, results in nonlinear differential algebraic systems. As computation time is a major issue, we are interested in applying model order reduction techniques to such multibody systems. One possible method called Proper Orthogonal Decomposition is based on minimizing the displacements euclidian distance while the more recently presented method Smooth Orthogonal Decomposition considers not only displacements but also their time derivatives. After a short introduction to the theory, this contribution presents a comparison of both methods on an index-reduced system. The methods are tested against each other in order to identify advantages and disadvantages.

In this study we present an adaptive STC controller design for small embedded systems by a new free tool-chain for model based control design. This is based among other software on the open simulator Scilab-XCos. After a very short introduction of model based design terms this article focuses on the code generator and the other programs of the tool-chain. The design concept is demonstrated by the non-trivial adaptive STC control of the cart system in simulation and on a real laboratory experiment.

In this paper we present free tools for model-based optimal input design and parameter estimation. The discussed tool-chain is tailored for the needs of small- and medium sized companies. Its programming core is based on Scilab and the JModelica platform and features input design (DOE), optimal control problems (OCP), and parameter estimation. Finally, the entire tool-chain concept is tested via simulation of a cart and pendulum system.

In this study we present a new free tool-chain for model based control design for mechatronic plants applicable to small embedded systems based among other software on the open simulator Scilab-XCos. After a very short introduction of model based design terms this article focuses on the code generator and the other programs of the tool-chain. The design concept is demonstrated by the non trivial adaptive selftuning control (STC) of the cart and pendulum system in gantry crane configuration in simulation and on a real laboratory experiment.

Often, trajectories for mechanical systems are generated solving some optimization problem. Common approaches include time-optimal, energy optimal, etc., motion profiles. In order to decrease mechanical wear of real plants this profiles provide, e.g., a smooth movement (rest-to-rest) in accordance with restrictions in jerk, acceleration and velocity. There exists a number of methods, to calculate for a given trajectory the plant feed forward action and to design stabilizing controllers. In case of parameter uncertainty the control law often exhibits some adaptive part. Unfortunately, smooth trajectories tend to contain insufficient excitation for adaption and/or identification. Therefore, we propose to consider some measure for the information content concerning some unknown parameters in the trajectory optimization problem.