Integration and test plans for complex manufacturing systems

The integration and test phases that are part of the development and manufacturing of complex manufacturing systems are costly and time consuming. As time-to-market is becoming increasingly important, it is crucial to keep these phases as short as possible, whilemaintaining system quality. This is especially true for the time-to-market driven semiconductor industry and for companies providing manufacturing systems to this industry such as ASML, a provider of lithographic systems. The Tangram research project has the goal, to shorten integration and test time by a model-based integration and test approach. The Ph.D. project described in this thesis is part of the Tangram project. To achieve integration and test time reduction, we developed three methods that each solve one of the following three integration and test problems: • Construction of an optimal test plan with respect to time, cost and/or quality. • Construction of an optimal integration plan with respect to time, cost and/or quality. • Construction of an optimal integration and test plan with respect to time, cost and/or quality. The test plan optimization method consists of two steps. The first step is the definition of a model of the test problem. This model consists of tests that can be performed with associated cost and duration, possible faults that can reside in the system with associated fault probability and impact (importance), and the relation between the tests and the possible faults, also denoted as the test coverage for each possible fault. The second step consists of calculating the optimal test plan based on this test model given an objective function and possible constraints on time, cost and/or risk, which is a parameter for the quality of the system. By constructing an AND/OR graph of the problem, where AND nodes denote tests and OR nodes denote system states represented by the ambiguous faults, all possible test sequences of this problem are obtained. An algorithm selects the best solution from this AND/OR graph. This solution is a set of test sequences, where the test sequence that is followed depends on the outcome (pass/fail) of the previous tests. The integration plan optimization method consists of the same two steps as the test plan optimization method. The integration model consists of modules with their development times, interfaces that denote which modules can be integrated with each other, and test phases with their durations. Furthermore, the model consists of the relation between test phases and modules indicating which modules should be integrated before the test phase may start. Also for this problem, an AND/OR graph is constructed. The AND nodes denote integration actions and the OR nodes denote system states represented by the modules that are integrated. An algorithm selects the optimal solution from this AND/OR graph. The optimal solution has the shortest possible integration time. The solution is a tree of integration actions and test phases indicating, for each module, the sequence of integration actions and test phases. The integration and test planning method is a combination of the two previously mentioned methods and also consists of two steps. The integration and test model is a combination of the test model and the integration model, with additional relations between modules and possible faults describing in which modules these possible faults are inserted. During the construction of the integration AND/OR graph, a test AND/OR graph is constructed for each integration AND node. This test AND/OR graph represents the test phase that is performed after that integration action. The start and stop moments of these test phases are determined by the test phase positioning strategy. We developed several test phase positioning strategies according to which test phases are started, for example periodically or when a certain risk level is reached. We applied the methods developed to industrial case studies in ASML to investigate the benefits of these methods. From a case study performed in the manufacturing of lithographic machines, we learned that the duration of a test phase may be reduced by approximately 20% when using the test plan optimization method instead of creating a test plan manually. From a case study performed in the integration phase of a new prototype system, we learned that using the integration planning method may reduce integration time by almost 10% compared to a manually created integration plan. From a case study performed in the integration and test phase of a software system, we learned that the final test phase durationmay be reduced by approximately 40% when applying a risk-based test phase positioning strategy instead of the currently used periodic test phase positioning strategy. We conclude that the methods developed can be used to construct optimal integration and test plans. These optimal integration and test plans are often more efficient than manually created plans, which reduces the time-to-market of a complex system while maintaining the same final system quality. Future research should indicate how to incorporate the methods developed in the complete integration and test process, and how to obtain the information needed to create the integration and test models.