Flexible Waveform Generation Accomplishes Safe Braking Just as the antilock braking sytem (ABS) has become a critical safety feature in automotive vehicles, it perhaps is even more important in railway systems where locked wheels can result in extensive damage to both the train and the tracks. Westinghouse Brakes (U.K.) Ltd. is a world leader in developing brakes for trains and metro systems. Originally founded by George Westinghouse more than a century ago, the company now is part of the Knorr-Bremse Group. Westinghouse has developed several Wheel Slide Protection (WSP) systems, which minimize stopping distance and prevent damage to the train and the rails. To aid in the design and testing of the WSP systems, Westinghouse Brakes engineers needed a flexible, yet affordable, way to generate arbitrary signals for a hardware-in-the-loop test system. The company turned to PCI analog output (AO) cards from United Electronic Industries (UEI) that could simultaneously generate multiple independent signals at frequencies up to 100 khz per channel. Different From Automobile Brakes Although the WSP system functions in a fashion similar to the ABS system in motor vehicles, there are some crucial differences. Both work on the same basic principle: the continuous modulation of the braking effort depending on the amount of wheel slip and adhesion conditions. However, in railway braking systems, the actuation is electropneumatic while an automotive ABS typically is electrohydraulically actuated. Other differences deal with adhesion and stability. In automotive braking, the
system must minimize any lateral movements and prevent loss of lateral stability of the vehicle, but that is less critical with a vehicle running on fixed rails. On the other hand, the amount of adhesion available to railway trains is inherently limited due to the extremely low steel-to-steel coefficient of friction between a train wheel and the track, and adhesion can further deteriorate in the presence of rain or foliage on the track. A well-designed WSP system not only must minimize the stopping distance by making optimal use of available adhesion, but it also must prevent any excessive sliding of the wheels on metal rails. This sliding generates high temperatures and can lead to modifications in the chemical structure of the steel, which cause flat points on the wheel. These deformations, in turn, produce mechanical damage to both the wheelsets and the tracks, resulting in higher running costs. For that reason, a WSP system must accurately control the amount of slip. Controlled railway-braking systems use an electropneumatic action working in a closed loop. Such a system measures wheel speed with a tachometer mounted on each axle. Depending on the amount of slip (the relative difference between train speed and axle speed), it issues a braking command to an electropneumatic control valve with incorporated electronics, which modulates pressure in the brake cylinders. It does so in a closed loop according to a control algorithm. If any part of this feedback chain fails, the system must enter a safe state.
Figure 1. Simulated Braking SessionA simulated braking session is depicted in Figure 1. Three axles are shown having a controlled amount of slip relative to the ground speed while the fourth is periodically released to allow calculation of the actual train speed. Railroad brakes have a long service life, with maintenance cycles every several years. They operate in the hostile under-train location with high vibration levels and extremely variable environmental conditions. For these reasons, reliability, robustness, and safety are key to the design of WSP systems. The capability to test their performance in a repeatable manner, in all possible operating conditions, and in the presence of failure modes is crucial to product development and safety assessment. But running comprehensive tests on actual trains is impractical. Such tests not only are highly expensive, but also are not repeatable because adhesion values change continuously along the track. Further, lab testing using purposely damaged or faulty equipment would not let designers assess system performance against all possible failure modes because the number of possible
electrical and mechanical faults and their combinations are high. Testing brakes in the lab requires, among other things, very accurate simulation of the control signal that represents axle speed. On a train, this measurement is made with an axle-mounted toothed geared wheel, called a phonic wheel, coupled to a magnetic sensor that generates a tacho signal consisting of pulses at a frequency proportional to the rotational speed. A host of undesired external factors can corrupt tacho signals. Among them are shock and vibration, misalignments, eccentricity, wear and the loss of teeth, and electromagnetic interference. To assess the performance and safety of products under these extreme conditions, engineers at Westinghouse Brakes developed a hardware-in-the-loop simulator dubbed Safety+ that can test brakes under both normal operating conditions and faulty conditions by injecting software-generated corrupted tacho signals. The system consists of the WSP coupled to a software simulator that models train motion. Actual measured pressure signals go from the WSP to the simulator, which returns a software-generated tacho signal to the WSP. In this manner, the system can react to various braking situations as if they were real. It records any deterioration in performance and checks results against a strict pass/fail criterion. Simulating the Tacho Signal One challenge in setting up this system was generating the simulated tacho signals. In an ideal world, it would be a variable-frequency rec-tangular pulse train of constant amplitude, which wouldn t be difficult to generate with a reasonably flexible digital I/O card with timing-sequencing features. However, the real feedback signal differs in many respects, and it actually resembles an arbitrary analog waveform. Specifically, the engineers had to augment a pure square wave to obtain corrupted signals to simulate the following conditions: Electrical interference. Power-supply variations.
Tooth broken off the phonic wheel or an eccentricity in its shape. Mechanical vibrations. Electromagnetic interference. Failure of the control electronics or damage in the tacho s magnetic circuit. In some circumstances, the tacho s output frequency can jump by a factor of 0.5x, 2x, or 3x, and the control system must detect such a jump and switch to the actual frequency. To generate signals that realistically emulate all possible mechanical and electrical fault conditions, the design team used a variety of signal-processing techniques. The resulting analog waveform is quite complex. Being able to generate a high-resolution waveform with a range from near DC to greater than 20 khz is beyond the capabilities of most AO cards. However, a product that could meet these specifications was UEI s PD2-AO-8/16. The test set generates four continuously variable-frequency/variable-amplitude square waves, representing the corrupted tacho signals, one corresponding to each axle in a train car. The most difficult part of the project was writing software to simulate the many possible error conditions that could be part of the tacho feedback signal. Each error modifies the software-generated pulse train in a subtle, yet critical, fashion, and the engineers had to duplicate these effects in software. They did so using special drivers that UEI supplied for its AO board. That card s onboard DSP proved crucial in solving the application requirements. Specifically, the AO modes on the card can stream data from a large datafile through the D/A converters and supply data from the DSP s onboard memory in a circular fashion without host intervention. When changing datasets, though, a very slight gap of perhaps a few samples can appear in the datastream. Westinghouse Brakes, on the other hand, needed a continuous stream of data, but this data consisted of a large number of small waveform segments. Further, the simulated tacho feedback signal could not have any artificial gaps or delays that the control system might construe as false failure modes and try to react accordingly.
To solve this problem, UEI s engineers developed a way to replace part of the data and a pointer within the DSP memory while the remainder of the memory was providing output data and do so without disrupting the output signal in any way. They also provided a specialized driver to support this new feature and supplied a C example for this specialized direct DSP control output mode. Although each AO card features eight independent 16-b analog outputs, the designers chose to use one output from one card for each axle. This requirement arose because WSP needed a separate clock for each D/A converter, and like almost all multichannel AO cards, this one uses a global clock for all the converters. One test of the WSP takes only several minutes, but a complete test suite encompasses many hours. This is because a complete test must take into account the interactions of many variables: whether the train is loaded or unloaded, different track-adhesion profiles such as dry or wet rails, the initial speed before braking, and the type of noise (white noise, jitter, or amplitude modulation of the tacho signal from different axles). Checking various combinations of these factors can lead to several hundred tests within a complete suite. The software runs through the suite automatically, and at the end, the Westinghouse Brakes engineers can provide a spreadsheet summarizing all the pass/fail results to the customer. Only when the WSP equipment has proven itself is it awarded the Safety+ logo and approved for release to the customer. About the Author Emanuele Guglielmino, Ph.D., holds an M.S. in electrical engineering from the University of Genoa and a doctorate in fluid power systems and control from the University of Bath. He worked for Westinghouse Brakes Ltd. as a control engineer on railway brake systems and most recently joined GE Energy in Florence, Italy. Dr. Guglielmino has authored more than 15 publications in the fields of robust control and fluid power and, in 2001, won the ASME Best Paper
Award in the Fluid Power Systems and Technology Division. e-mail: e_guglielmino@yahoo.it