Research on Real-time Performance and Reliability of CAN Bus
|Course||Power Electronics and Power Drives|
|Keywords||CAN bus Response time Queueing theory Time Series analysis Reliability Test system for response time|
CAN Bus is widely used for its low cost, high performance and flexibility. Although CAN bus is adopted in safety-critical systems such as automobile and small aircraft, many scholars and engineers focus on how to evaluate the real-time performance and reliability of CAN-bus either by theory or practice and try to show that CAN bus is able to be utilized in the applications which are strict with the real-time performance and reliability.Queueing theory evaluates the performance such as mean value when the system is in balance. According to the CSMA/CD protocol of CAN bus, a nonpreemptive priority M/G/1 model was presented, and the mean value of response time was educed. Furthermore, the relationship between the mean value of response time and bus load was given. The above results can help the design and optimization of CAN based network system greatly.An implementation of time/event-triggered state machine was given and Worst Case Response Time Analysis was presented for the mixed-triggered state machine. Worst Case Response Time Analysis doesn’t take the local blocking into account, therefore an analysis of Worst Case Response Time considering hardware and software buffer and the schedulability was proposed. The presented method exercised a message dataset of electric vehicle.Because there’s no instrument to measure the response time of messages of CAN bus, a test system for the measurement of response time was developed. Synchronization of clocks based on GPS, Phase-locked loop and the share of real-time clock based on the semaphore with priority was used. The experiment results showed that the precision was high enough for the research on the statistical analysis of response time.By means of time series, the experiment data was analyzed and the result showed that the model of response time varied with the different message dataset. Some dataset was described by ARMA model while others was suitable for the long memory model. The conclusion provided a new approach to the analysis of response time.Based on the proof that probability of successful transmission should not be used to analyze the reliability of message dataset, artificial neural network was induced to simulation to calculate the reliability, which reduced consumption of both data and time when the simulation was only utilized for the calculation of reliability over long time. To meet the requirement of the reliability engineering, the electrical fast transient/burst interference that conforms to the international and national standard was used in the above method.