Array based signal processing is a method for identifying important noise sources. Different array configurations have been used in the past. Random arrays are thought to have better resolution and precision for identifying moving noise sources than linear or cross arrays. But how to design optimal random configurations is a challenge. Based on property analysis and simulation of random arrays with different configurations, evaluation criteria for parameters of two-dimensional random arrays are proposed, and a random array generation method to locate the noise sources on moving vehicles is presented. It consists of an array unit partition, baseline filter and simulation based evaluation. The performance of the generated array is shown by identifying noise sources on a passenger car.