Computed Expert System of Support Technology Tests in the Process of Investment Casting Elements of Aircraft Engines Krzysztof Zaba 1 *, Stanislaw Nowak 1, Adam Sury 2, Marek Wojtas 3, Boguslaw Swiatek 1, Rafal Cygan 4, Michal Kwiatkowski 1, Maciej Nowosielski 1 1 AGH University of Science and Technology, Faculty of Non-Ferrous Metals, al. A. Mickiewicza 30, 30-059 Krakow 2 Infoster Sp. z o.o., Krakow, Poland 3 Elsta Elektronika Sp. z o.o. S.K.A., Krakow, Poland 4 WSK PZL Rzeszow S.A., Rzeszow, Poland e-mail: krzyzaba@agh.edu.pl The paper deals with the manufacturing jet engine turbocharger blades, made of heat-resistant and creep-resisting nickel and cobalt superalloys. Previous studies and manufacturing experience has shown that current, widely used technology of precision casting using the lost material (eg. wax) does not provide the conditions for obtaining casts a fully consistent quality, described by the shape, dimensions, surface condition, edge condition and the size and distribution of defects reasonable at level of internal economic considerations. The reason is that the production process is extremely complex process in which many material and technological and organizational factors influences the final quality. In this situation, a two-way action is being taken. The first direction is to improve the operational monitoring and control, aimed at stabilizing the processes by eliminating the emergency causes of variation. The second - the improvement of technology and processes, which require, among other, planning, implementation and tests of technological and appropriately targeted inference. Research may be done either in a line or beyond. In this series of activities are also part of projects consistent with the methodology of Six Sigma. Keywords: technological tests, data acquisition, data analysis, Introduction The paper presents the conception and an example of implementation, and use of the Computed Expert System of Support Technology Tests in the process of investment casting elements of aircraft engines. Support includes the design, monitoring, analysis of results and inference. Investment casting method in multi-layer ceramic molds makes possibility to reproduce geometrical shapes of aircraft parts, such as blades, vane clusters of the gas flow turbine wheel, housing and others. Figure 1 shows different vane clusters and blades. The manufacturing process of blades, is a complex process in which many material, technological and organizational factors affect on the final quality. Main problem is that the quality is evaluated post factum. This is due to the lack of an effective form of quality control methods in the next stages of its production. Figure 1 Vane clusters and blades Molds are made in several stages by layering the wax pattern, drying, melting wax and burning and annealing. Casting takes place in a vacuum furnace. Each of these phases is susceptible to disturbance. In addition, there is no effective methods for inter-operational control, allowing to specify the state of the mold and its early elimination. In this situation, a two-way action is being taken. The first direction is to improve the operational monitoring and control, aimed at stabilizing the processes by eliminating the
emergency causes of variation. The second - the improvement of technology and processes, which require, among others, planning, implementation and tests of technological and appropriately targeted inference. Research may be done either in a line or beyond. In this series of activities are also part of projects consistent with the methodology of Six Sigma. The paper presents the concepts of computer system, aimed at supporting these activities. The system has 7 basic functionalites. 1.Creating, updating and sharing Knowledge Base 2. Database creating which characterizes primary and secondary. feedstocks 3.Creating base technological problems, including 4. Database creating and sharing data on implemented internal and external projects 5.Supporting in designing technological tests using tools for planning experiments and the Six Sigma methodology, including the identification 6.Monitoring of technological tests 7.Analysis of performance and powering interference Scheme of Computed Expert System of Support Technology Tests shows Figure 2. manually-entered process-related data Operator inferencing & analysis module Operator Operator Operator MATERIALS PROJECTS PROBLEMS KONWLEDGE P R O C E S S SERVERS real-time-visualisation modules, data acquisition DATA ACQUSITION modelling & statistics planning experiment Operator Operator Operator Operator Operator Operator Operator LABORATORY TEST STANDS Figure 2 Scheme of Computed Expert System of Support Technology Tests The system can be divided into three logical layers. The first concerns the process data acquisition from the trials carried out in technology line or from test laboratory tests - out of the technology line. The second concerns the planning of experiments, their monitoring, data analysis and inference. The third is responsible for archiving data. Verification of System was made during launching of a new product. Purpose of tests was to collect of experimental material, which is the basis for the improvement of mathematical models of processes. Studies have also indicated the parameters necessary for measuring and recording to complex processes monitoring. An important element of the system was to identify the unit, which was a single blade. Identification of the blade is provided in the subsequent process, from the manufacture of the wax pattern by positioning the mold in the process of melting the wax, annealing and casting. In this way obtained the possibility to compare the results of the quality with the parameters setted for the individual process. The quality of the blades was evaluated on the basis of visual inspection, shape and size inspection by using a dedicated measurement systems, the results of FPI and X-ray results. During the tests were recorded: -Identifiers, -Process-execution parameters, -Time, -Service. Data analysis module provides the processing of data to the expected form, the choice of main information such as the maximum value of the measurement results in the time decomposition of some feature, visualization, statistical modeling, matching the coefficients of physico-chemical models and the construction of models based on regression and correlation analysis. System conception The system uses the existing IT infrastructure and logistics business and achieves three criteria: 1. Flexibility. 2. A maximum modularity. 3. Technology client-server. The criterion of flexibility can implement system into the existing structure of IT structures and download information from the traditional storage systems, materials management, finance and accounting, and
taking orders. There is a problem with adapting interfaces, to make ability to import and export data to and from the system. During import and export, if needed, the data is automatically converted to the required form. Maximum modularity criterion leads to break the problem into small tasks. Each of them have been parameterized and programmed. Thus in a specialized, small module is solving a specific task. This approach (modularity) provides easy expansion of the system and adapt to the specific needs and increases reliability, simplifies modification and maintenance. In addition, modules can be implemented in stages, as far as solving the following problems. Cohesive element creating system is Database System. The functional structure of the system is shown in Figure 3. Substance & form data sharing, Analysis and visualization of the data contained in the projects, Inference on the efficiency and effectiveness of the solution to the problem, development, prospects for cooperation, Evaluation of co-operating units, Assessment teams. P R O C E S S measurement modules manually-entered process-related data Acquisition of measurement data, The creation of film and photographic documentation, Creating documentation of tools, Creation of documentation of material test results, Creating documentation of modeling and computer simulations. Feedstock Primary Secondary MATERIALS Projects RealTime Data Server Problems Knowledge reporting module Descriptions of problems, IIdentification of needs, Expected results, Risk, History of the problem Materials, Tools, Equipment, Requirements, standards Criteria and constraints Models inferencing & analysis module real-time-visualisation modules modelling & statistics Figure 3 Functional structure of the System The system has 7 basic functionalities. 1.Creating, updating and sharing Knowledge Base, on: -Materials, tools, equipment, -Requirements, standards, -Criteria and constraints, -Models. 2. Database creating wchich characterizes primary and secondary. feedstocks 3.Creating base technological problems, including: -Descriptions of problems, -Identification of needs, -Expected results, -Risk, -History of the problem. 4. Database creating and sharing data on implemented internal and external projects -Data-sharing substantive and formal -Analysis and visualization of the data contained in the projects, -Inference on the efficiency and effectiveness of the solution to the problem, development, prospects for cooperation, -Evaluation of co-operating units, -Assessment teams. 5.Supporting in designing technological tests using tools for planning experiments and the Six Sigma methodology, including the identification: -Methodology, -The necessary resources, -Cost, -Effectiveness. 6.Monitoring of technological tests: -Acquisition of measurement data, -The creation of film and photographic documentation, -Creating documentation of tools, -Creation of documentation of material test results, -Creating documentation of modeling and computer simulations. 7.Analysis of performance and powering interference: -Analysis of trials are ongoing, aimed at assessing the efficiency and effectiveness of the tested variant, -Analysis of historical data,, Tables of Constant Values (TCV) perform the function of System integration. Included in these dictionaries, norms, structures are supported (maintained and verified) in accordance with the rules of access. With TCV electronic version of the organizational technical and cost documentation, is maintained. TCV is built according to the requirements of users and provide different access rights to change the structure of the data, full control of data integrity, the development of the presentation, flexible of module to communicate with other programs.
The tables are maintained and verified by special rules. Verifying data is in accordance with specially developed procedures. Module data analysis and inference tool is equipped with: - Analysis based on models of phenomena occurring in the processes (physical, chemical and created on the basis of experimental data) - Enabling the inference rules in the database. The example involves testing technology for manufacturing vane clusters of jet engine turbine. During the test process data is automatically recorded in the data acquisition module and harvested by hand when one of the selected measurement parameter was not included in the system. Sample results are shown in Figure 4-13. Figure 4 shows the variation in the time of injection wax to die during a single wax patterns. Figure 6 Degreasing time in solutions of model sets 1, 2, 3, 4 Figure 7 shows the results of the identification process, the viscosity of the mixture, coating I.. Figure 7 Viscosity of the mixture in the process of the ceramic coating 1st Figure 8 shows the results of the identification process, the ph of the mixture coating 1 st. Figure 4 Time of injection wax Figure 5 shows the results of the identification of the temperature in the tank stabilizing wax during wax patterns of individuals. Figure 8 ph in the mixture of the ceramic coating 1st Figure 9 shows the results of the identification of sets of time of immersion in a tank with a mixture of the ceramic coating Figure 5 Wax temperature in stabilization tank Figure 6 shows examples of the results of identification time degreasing wax model sets of solutions 1, 2, 3 and 4 Figure 9 The time of immersion in the reservoir model sets with a mixture of ceramic
Figure 10 shows the results of the identification time the mixture dripping ceramic coating I. Figure 10 Time dripping of ceramic mixtures Figure 11 shows the results of the identification time III ceramic coating by a robot. Figure 11 Time of coating ceramic layer III Figure 12 shows the results of the identification of the humidity and temperature of the hand cover facility. Figure 12 Humidity and temperature of the hand cover facility Figure 13 shows an example of climate variability in the facility to cover the hand. Figure 13 An example of the variability of temperature and humidity of the hand cover facility Example of analysis with usage of neural networks Analysis and inference module is equipped with a tool to build models that are created based on experimental results using a neural network methodology. Solved are two types of tasks. 1. Determination optimal set of parameters implementation, ensuring the fulfillment of the assumed objective function. In this case, minimizing the disadvantages. 2. Approximation of sought functions, in principle, a number of variables using experimental data. To solve these problems special program was built. All stages of technology, combining unit processes in a group, variables, and modeling compounds of these variables are defined by the user. The models are stored, you can improve by introducing new data portion. The calculation results are presented in numerical and graphical form. The measurement data are entered into the database regardless of the design procedures and defining models. Models can be created for existing data in a variety of (any) configurations. Network parameters are determined in the learning stage based on the input results. Learning networks after each batch of experimental data and in this sense the program is adaptive in nature. After defining of model, choosing of right data from Database and program of learning the neural networks is launched. After finishing learning structure of network is remembered at HDD such as models name. Acces is possible after typing right name. Artificial neural networks (ANN) are effective tools for solving this type of problems mainly due to their ability of approximation of any multidimensional non-linear function. The approximated function is obtained in the network training process, and this feature singles it out from other technical systems. Network training is based on presenting the training set it means the set of the process parameters values and the product parameters values. During the training process the network
modifies its parameters in such a way as to find their relations. The result is in a form of the network model, being simultaneously the process model. The network excitation by values corresponding to process parameters causes that values corresponding to the product parameters occur at the output. The problem is presented schematically in Figure 14. Figure 14 Processing of input data into output data by means of the neural network Conclusions Management of technical documentation, documents, messages and all kinds of materials that can be stored in electronic form, its using groupware tools. It allows to systematize all kinds of business processes and workflows with standard intranet portals and dedicated applications. Electronic materials are stored in data repositories. Its providing group and versioning. Model circulation of documents and information based on a many to many relationships, which means flexible linking information from the users. Users and documents are organized into groups with defined permissions and access rights. Group members are: - administrators - users with the ability to create and modify documents - users of the possibility of modifying documents - users of the possibility of viewing documents group of documents are: - photographic and film documentation - documentation tools - documentation of the materials research - documentation of computer modeling and simulation Process groups are: - projects - research - technological tests - management of the technical documentation Groupware tool also provides high security data and control, in conjunction with the mechanisms of operating systems such as Windows, permissions and access rights. User interfaces are designed as standard portals supported by web browsers and applications that work with SharePoint platform. System uses a SharePoint database and dedicated database designed to store information about processes (technological tests, research). References [1] Korbicz J., Obuchowicz A., Uciński D.; Sztuczne sieci neuronowe, Warszawa, 1994. [2] Russell S.J., Norvig P., Artificial Intelligence A Modern Approach (3rd Edition), Prentice-Hall, 2009. [3] Akerkar R., Sajja P., Knowledge-Based Systems, Jones and Bartlett Publishers, 2009. [4] Giarratano J.C., P. Gary, Riley D., Expert Systems: Principles and Programming, (4rd Edition), Thomson Press, 2004. [5] Beynon-Davies P., Inżynieria systemów informacyjnych, Wydawnictwa Naukowo-Techniczne, 1999. [6] Osowski S.; Sieci neuronowe w ujęciu algorytmicznym, WNT,Warszawa 1996. [7] Żurada J., Barski M., Jędruch W.; Sztuczne sieci neuronowe, PWN, Warszawa 1996