Diagnosis Using Support Vector Machines (SVM)
Diagnosis of functional failures at the board level is critical for improving product yield and reducing manufacturing cost. State-of-the-art board-level diagnostic software is unable to cope with high complexity and ever-increasing clock frequencies, and the identification of the root cause of failure on a board is a major problem today. Ambiguous or incorrect repair suggestions lead to long debug times and even wrong repair actions, which significantly increase the repair cost and adversely impacts yield.
In this chapter, we introduce a machine learning-based intelligent diagnosis system, which can automatically learn debug knowledge from empirical data and identify the most likely root cause of a new failed board. Using such a diagnosis system eliminates the difficulties involved in traditional knowledge acquisition. Fine-grained fault syndromes extracted from failure logs and the corresponding repair actions are used to train the system. Support vector machines (SVMs) have been used in board- level diagnosis to provide accurate root cause isolation. An SVM-based diagnosis system can be rapidly trained and is scalable to large datasets. However, the SVM method used in prior work [1] was simplistic, relying on an arbitrarily chosen kernel function, and it was not adaptive to the availability of new data or test cases. We propose a diagnosis system based on multi-kernel support vector machines (MK-SVMs) and incremental learning, which are used to tune the diagnosis system in an automatic manner. The MK-SVM method leverages a linear combination of single kernels to achieve accurate faulty component classification based on the errors observed. The MK-SVMs thus generated can also be updated based on incremental learning, which allows the diagnosis system to quickly adapt to new error observations and provide even more accurate fault diagnosis.
The remainder of this chapter is organized as follows. Section 2.1 reviews the background and prior work. Section2.2 reviews basic concepts in support vector machines. Section 2.3 introduces multi-kernel-based SVMs, and describes how MK- SVMs can be extended for incremental learning, namely iMK-SVMs. Section2.4 presents experimental results on diagnosis accuracy and training time for two industry boards and for synthetic data. These results are compared to diagnosis using singlekernel SVMs [1] and ANNs [2]. In addition, experimental results are presented for the diagnosis accuracy achieved using incremental learning. The high diagnosis accuracy, rapid training, and short diagnosis time highlight the benefits of the iMK- SVM-based reasoning system. Section 2.5 concludes the chapter.
