Expert Systems in foundry industries Adam Sury Infoster Sp. z o.o. asury@infoster.com.pl Marek Wojtas Infoster Sp. z o.o. mwojtas@infoster.com.pl Michal Kwiatkowski Faculty of Non-Ferrous Metals AGH University of Science and Technology Krzysztof Zaba Faculty of Non-Ferrous Metals AGH University of Science and Technology krzyzaba@agh.edu.pl Rafal Cygan Investment Foundry W-50 Pratt & Whitney Rzeszow rafal.cygan@wskrz.com * Michal Kwiatkowski Faculty of Non-Ferrous Metals AGH University of Science and Technology ABSTRACT The paper shows method of design and build an expert system focused on special processes occurring in the production of aircraft engine parts manufactured by the investment casting. The construction of an expert system for technological processes in the foundry industries is determined by the class of problems. The way the complexity of the process, a multitude of physical and chemical reactions practically eliminate possibility to create a classical mathematical model for group of these processes. Knowledge base of the shown system is continuously updated collection of process parameters, with the greatest possible number of technological nodes. The first element of the system is a classic data acquisition system based on SCADA and fast relational DB. The system software performing functions of data analysis module and the requesting machine is based on numerical algorithms of data statistical processing and neural networks. "Data presentation and analysis" module are performed advisory functions for expert-technologist. An important element of the system is also dedicated to the production unit tracking system, which data are collected automatically supplemented by the information entered manually by the operator. The purpose of this system is to reduce product defects occurring during the investment casting process. Designed system can be classified as a dedicated advisory and diagnostic system with the uncertain knowledge. Keywords: expert system, foundry industries, data acquisition, neural networks 1 INTRODUCTION The history of expert systems is related to the development of scientific fields dealing with problems of smeared logic, neural networks, robotics, and process control. The scope of domains with which these systems are dealing is very broad from chess play via pictures recognition for medical diagnostics. In the years 70 to 90 of the last century great hopes were pinned to systems using artificial intelligence and smeared logic methods. Unfortunately there is a stagnation in the development of these domains and it is difficult to expect any significant turning point. Excellent
algorithms were developed used in individual applications, e.g. algorithms for sound pattern recognition, medical diagnostics, control, ect. however saying that this is already the artificial intelligence is too much. Expert systems used in diagnostics and in control of complex technological processes apply methods of classic artificial intelligence however directed towards a close collaboration with a man, who is an expert in the given field and serve rather as assisting systems and supporting diagnostics. Three criteria of dividing expert systems can be assumed. These are: interaction with a man, the way of realisation and the way of operation. 1.Interaction with a man: - closed systems, it means undertaking the decision without any ingerence and inspection by a man, applied in controlling complex devices under conditions which are excluding or highly limiting a man s possibility, - open systems (assisting) it means suggesting the solution helpful in undertaking the decision by a man. They present the solution of a certain problem but the user has to assess its value and decide whether he will accept or reject this solution. - critical systems, it means performing the analysis of a certain problem and its solution and then commenting the proposed solution. 2. The way of realisation - dedicated systems created by the science engineer together with the computer expert with the full science data base defined at the system designing stage, and this science base can be supplemented during the system exploitation, - skeletal systems with an empty data base defined during the system exploitation in the training process. 3. The way of operation - real time systems processing incoming information e.g. from industrial automatic systems, they acquire real time expertise, - analytical systems acquired expertise is not required in exact time limits. The remaining criteria of dividing the expert systems are related to their structure, computational methods, applied algorithms and the way of conclusions drawing. Such divisions are not accurate and the real systems in the majority of cases - use various architectures and ways of concluding. Each expert system is composed of three basic elements: 1. Science data base set of information concerning the problem being the subject of the system operations. These information are divided into rules and facts, which after being processed by the system lead to the expertise development. 2. Inferring machine set of algorithms (computer programs), which are to process information contained in the data base together with the real time information entering the system e.g. from measuring equipment of industrial automatics and from the system operator (employee). 3. User interface used for communication between the system and user. 2 EXAMPLE EXPERT SYSTEM IN FOUNDRY INDUSTRIES Lab of Production Engineering Team implemented so far several dedicated expert systems for in metallurgical, aerospace and petrochemical industry. Several of implementation included foundry processes. The paper presents an example of an expert system, built in the investment foundry Pratt & Whitney Rzeszow. The main task of this System is supporting of the ACE methodology (Achieving Competitive Excellence), which is held by UTC Corporation (United Technologies Corporation). Particularly this support provides: -non-conforming products quantity minimizing, -minimize of the instability of process both in their own variability and extraordinary variability, -cause and effect analysis, which provides preventive and corrective actions, -the withdrawal of the improper blank in an early production stage, -improvement of the OEE indicators.
Fig. 1. shows the location of an expert system among business systems in the enterprise APS advanced planning system planning, simulation, optimization strategic management Capacity Requirements Planning (CRP) Material Requirements Planning (MRP) work stations enterprise resource planning (ERP) tracking, visualization, reporting reporting system Manufacturing Execution System (MES) Supervisory Control And Data Acquisition (SCADA) servers dedicated systems Expert System statistics, analysis, neural networks, scheduling industrial automation equipment production process Fig. 1: Expert System structure 2.1 Basic tasks The presented here system realises 11 tasks: 1. Collecting of reliable data in the real time with the elimination of errors and delays introduced by a man process monitoring and the results of interoperational and final inspection. 2. Collecting feedback information from processes of further processing and exploitation data. 3. Analysis, data visualisation and assisting operational decisions with the usage of statistic methods, SPC. 4. Modelling and simulation of processes directed to the product quality forecasting. 5. Simulation of processes directed towards planning of the order book realisation and detailed scheduling. 6. Process control, directed towards parameter stabilizations, eliminating extraordinary variations and decreasing standard variations. 7. Control of state and status of equipment, devices and measuring systems. 8. Determination of the key characteristics of processes. 9. Analysis of archive data and drawing conclusions for process improvements. 10. Development and transmitting automatic warning signals. 11. Building science base, on the basis of archive data, necessary for creating mechanisms of drawing conclusions, which will support the decision processes. 2.2 Realisation The example of the System realisation was directed to ensuring the stability of unit processes parameters, development of corrective signals, looking for reasons of defects and improvement of production processes of aviation parts obtained by the investment casting method. The pictorial diagram of the system is presented in Fig. 2.
Process 10 Operators workstation Testing workstation Engineering workstation HISTORIAN Server s Expert DB Server s REDUNDANCY Tracking & Genealogy workstation SCADA Server SCADA Server Process 1 Process 2 Process 9 Process 5 Process 6 Process 11 Process 13 Process 14 Process 3 Process 7 Process 8 Process 4 Process 12 Fig. 2: Pictorial diagram of the system The logic of the information flow in the system together with information protocols is shown in Fig. 3. equipment protocols SCADA Server databases wax injection machine KepServer SNP wax injection machine KepServer SNP wax injection machine DASOMFINSEnet Producer Firing furnace Burnout furnace Autoclave Vacuum furnace Vacuum furnace Heat treatment furnace Heat treatment furnace DASGESRTP KepServer MODBUS TCP KepServer MODBUS TCP KepServer MODBUS TCP DASMBSerial MODBUS Serial RSLinx OPC Server KepServer DF1 KepServer DF1 APPLICATION SERVER DATA ACQUISITION HISTORIAN DB Burnout furnace DASMBSerial MODBUS Serial Expert DB Air-conditioning Merz.OPC_SAIA_S-BUS.1 Air-conditioning Merz.OPC_SAIA_S-BUS.1 Fig. 3: Logic of the information flow in the system Actualisations and modification in the system are performed with the proper safety standards (Fig. 4). Because of that all actualization and modification operations are carried out first in the test environment and only after performing all tests they are implemented in the production system. runtime database BACKUP archival database DATABASE AGH EXPERT Workstation Workstation Client Workstation DATABASE AGH EXPERT QA test database Test Workstation Fig. 4: Actualisation and modification policy
2.3 Example functionalities The results of various analyses performed by means of programs included in the Expert System are shown in Fig. 5-9. Fig. 5: Example of the control card form balance Fig. 6: Detailed analysis of the given series Fig. 7: Analysis of the ceramic mould mass results
Fig. 8: Individual mould localization in single operation Fig. 9: Defects localization for single blade 3 CONCLUSION 1. The system provides access to reliable data in real time, eliminating errors and delays generated by humans. 2. The system provides access to feedback from downstream processing and operational data and to establish an effective mechanism for the use of data. 3. The system supports the inference and decision-making based on the results of the analysis and knowledge base. 4. The system allows you to control processes aimed at stabilizing parameters, eliminating or reducing of variability. 5. The system provides monitoring the condition and status of equipment, devices and measurement systems. 6. The system allows the inference and support decision-making at all levels of processes management based on the knowledge base created using archival data. REFERENCES [1] Internal materials of Lab of Production Engineering (reports of R&D projects, expertises realised for Pratt & Whitney Rzeszow). [2] Internal materials of Pratt & Whitney Rzeszow.