1 Introduction
With the increasing number of high-rise buildings in modern cities, people have put forward higher requirements for vertical transportation systems, namely, to improve the transportation efficiency of high-rise buildings, reduce the land area, save energy and reduce construction costs. Single-car elevator systems can hardly meet the needs of high-rise buildings with large passenger flow transportation[1]. People have begun to study increasing the number of elevators in the shaft based on the original single elevator car in one elevator shaft, i.e., multi-car elevators[2-3]. The problem that follows is to prevent collisions between multiple cars. Therefore, the problem of preventing collisions in multi-car control systems has become one of the hot research topics of scholars today[4].
Since multi-car elevator systems are still in their infancy internationally, research on collision prevention for multi-car elevators is not yet mature. Reference [5] uses sensors to measure the distance between two adjacent cars. When this distance is less than half the car body length, a collision is judged to have occurred, causing one of the cars to decelerate. This method controls the movement of the car from the hardware perspective, which has poor timeliness and a certain false judgment rate. Reference [6] uses the EVALPSN safety verification method to apply to the collision prevention of circulating multi-car elevators, but there is still no solution for collision prevention of single-shaft multi-car elevators. The focus of this paper is to mathematically model the motion law of the two cars and conduct theoretical research on the system model. For single-shaft double-car systems, due to the complexity and fuzziness of the system input variables, a radial basis function neural network algorithm [7-8] and a control method based on the safety judgment criterion are proposed to effectively prevent car collisions.
2. Multi-car safety distance model
2.1 Kinematic Description of Multi-Car Units
The basic composition and function of the single-shaft dual-car anti-collision system are shown in Figure 1. The hardware system detects the car's operating environment and obtains the car's basic information. The anti-collision system needs to identify the influence of each state parameter on the car's operation. The detected variables are used as inputs, and the corresponding relative relationships are obtained through data processing [9]. Finally, the actuator takes corresponding measures, such as alarm and braking, based on the processing results.
Figure 1. Composition of the anti-collision system for multi-car elevators
Figure 2 Single shaft multi-car operation
Figure 3. Collision prevention process for single shaft double car.
Figure 4. Relationship between RBF training error and number of neurons
Figure 5. RBF training error graph
Based on the sample data of the input vector, 16 sets of data were randomly selected for testing, and then the original function was input and compared with them to obtain simulation figures 6 and 7.
Figure 6 shows the collision probability predicted by the algorithm in this paper.
Figure 7 shows the error between the predicted and actual values of RBF.
5. Conclusion
To address the collision avoidance problem in multi-car elevators, a collision avoidance mathematical model based on an RBF neural network is proposed. By designing and determining several key parameters during the modeling process, and combining them with safety judgment criteria, a collision-free safe operation of the multi-car elevator can be guaranteed. Simulation experiments conducted in the MATLAB environment show the effectiveness of the RBF neural network, demonstrating its ability to implement braking and alarm measures based on different safety values. However, due to the complexity of multi-car systems and the influence of objective factors, further optimization of specific experiments is needed to provide higher assurance for the operation of multi-car elevators.
