A software fault refers to a mistake or imperfection in the coding of a program that results in malfunctions or inaccurate output. It usually results from development errors and might cause problems with the system or other problems. Software fault localization requires identifying the precise location of defects within a program’s code to facilitate efficient debugging. It involves techniques such as automated testing, debugging tools, and fault isolation methods. By using algorithms to evaluate code, find trends, and rank potentially problematic locations, machine learning helps localise software faults more quickly and more accurately. This paper employs the Siemens program suite from SIR to evaluate our proposed model effectively. Benchmark programs are executed with varied test cases, resulting in pass or fail outcomes. Machine learning models are trained using the coverage matrix, generating suspiciousness scores for statements to expedite fault identification and minimize debugging time. We evaluated the accuracy of our proposed model using the EXAM metric. Based on the results, the CNN model demonstrates superior performance, surpassing the RNN by 3% and the ANN by 6%. Due to its enhanced performance, the CNN model is recommended for reducing fault detection time.