The single fault and multiple fault detections for multiple-valued logic circuits are studied in this paper. Firstly, it is shown that the cardinality of optimal single fault test set for fanout-free m-valued circuits with n primary inputs is not more than n + 1, for linear tree circuits is two, and for multiplication modulo circuits is two if n is an odd number or if n is an even number and m > 3, where the optimal test set of a circuit has minimal number of test vectors. Secondly,it is indicated that the cardinality of optimal multiple fault test set for linear tree circuits with n primary inputs is 1 + [n/(m - 1)], for multiplication modulo circuits is n+ 1, for fanout-free circuits that consist of 2-input linear tree circuits and 2-input multiplication modulo circuits is not greater than n+ 1, where [x] denotes the smallest integer greater than or equal to x. Finally,the single fault location approaches of linear tree circuits and multiplication modulo circuits are presented, and all faults in the two types of circuits can be located by using a test set with n + 1 vectors.
The circuit testable realizations of multiple-valued functions are studied in this letter. First of all,it is shown that one vector detects all skew faults in multiplication modulo circuits or in addi-tion modulo circuits,and n+1 vectors detect all skew faults in the circuit realization of multiple-valued functions with n inputs. Secondly,min(max) bridging fault test sets with n+2 vectors are pre-sented for the circuit realizations of multiple-valued logic functions. Finally,a tree structure is used instead of cascade structure to reduce the delay in the circuit realization,it is shown that three vec-tors are sufficient to detect all single stuck-at faults in the tree structure realization of multiple-valued logic functions.
The main task of system reliability design is to find the best layout of components to maximize reliability or to minimize cost. A reliability optimization approach using neural networks to identify the choice of components in series-parallel systems with multiple constraints is presented in this paper. The McCullochPittes neural network model is used in this approach. The design methods of the neural network construction and its energy function are described in detail. The optimal solutions of the reliability problem are obtained by minimizing the energy function of the neural networks. Simulation results show the reliability optimization approach using neural networks can find the optimal or near-optimal solutions for most of the problems in a relatively short time, it is a useful alternative for system reliability design of complex systems.