Fuzzy logic implementation of an electromagnetic interactions modelling tool

An electromagnetic interactions modelling tool which is based on a fuzzy logic representation of the electromagnetic attributes in a topological decomposition of a system is described. The purpose of this tool is to help determine any electromagnetic compatibility problems in complex systems. This tool is an extension of the HardSyslHardDraw software [l, 21 enabling it to handle a fuzzy representation of the electromagnetic interaction data. HardSys, a prototype system implemented in Prolog, is weld to propagate the electromagnetic information through the topology of the represented system. User interaction is through HardDraw, ,an electromagnetic topology drawing tool and an attribute interface. Introduction The adverse effects of electromagnetic interactions in electrical systems are of concern because of the increased pollution of the environment with electromagnetic emissions and because of the increasing susceptibility of system components. From a practical point of view, it is not a simple matter to ensure {he electromagnetic integrity of systems even for relatively small interaction problems. Non-algorithmic or heuristic techniques are used daily by engineers to solve electromagnetic problems in electrical systems. An attempt to formalize these procedures in the form of a computer tool called HardSys/HardDraw was described in [l, 21. The modification of the knowledge representation used in this prototype tool into a fuzzy form [3] is described, This allows the heuristics and uncertain information associated with an interaction problem to be modelled more realistically than was possible in the first version of the tool. Electromaanetic T o r > o l w o f s t t r m s The electromagnetically relevant attributes of an electrical system can be isolated by decomposing the system into its corresponding electromagnetic shielding topology and its dual graph or interaction sequence diagram [4 , 5 , 61. The electromagnetic topology consists of a description of the electromagnetically distinc t volumes and their associated surfaces. The volumes define the electromagnetic components involved in the interaction. The interaction sequence diagram keeps track of the interaction paths throughout the system. The interaction sequence diagram can be simply derived from a given electromagnetic topology. The graph representing a simplified topology of a computer is shown in Fig. 1. Note the different node representation for field nodes, circuit nodes and interaction path nodes [ 1,2]. William H. Henneker Knowledge Systems Laboratory Institute for Information Technology National Research Council Ottawa, Ontario, Canada, K1A OR8 e-mail: bill@ai.iit.nrc.ca Power Cab EM1 Filter Circuit Electronic Distribution Circuitry Interaction Circuit PathNode Node Fig. 1. Interaction Sequence Diagram for a Simple Topology Interaction path nodes, or simply surfaces, are of four types: ffnodes, @-nodes, cf-nodes and cc-nodes. These distinguish between paths connecting the different combinations of field nodes and circuit nodes. The specific type of surface node will determine the type of attribute required to approximate the propagation of energy across that surface. Electromaanetic Attributes The next step in modelling the electromagnetic system is to approximate the propagation of electromagnetic energy from one volume node to another. Fuzzy electromagnetic attributes are introduced for each electromagnetic component in the topology as well as for the interaction paths between the components. These attributes approximate the propagation of the electromagnetic disturbances throughout the topology and represent the electromagnetic knowledge which is known about a system. Each volume node in an electromagnetic topology may have one or more electromagnetic disturbances (D) associated with it. These disturbances are represented as fuzzy variables with trapezoidal membership functions [3] as shown in Fig. 2 below. An important property of the trapezoidal functions is that they can be represented by the 4-tuple (a , b, c, d) with a 5 b 5 c 5 d. The meaning of a designation such as [(IO, 20) MHz, (10, 15, 20, 22) dBmV/m/Hz] could be translated as: " in the frequency range of (10, 20) MHz the electric field of this disturbance has a good possibility of lying between 15 and 20 dBmV/m/Hz but can be as low as 10 dBmV/dHz and as high as 22 dBmV/m/Hz ' I . An entry for a specific disturbance, such as circuit board radiation, is made up of a list of entries such as this which fully cover the required frequency range. For example, the emissions from a circuit board may be defined as (hypothetical): CH3169-0/9210000-0023 $3.00 01992 IEEE 127 (disturbance, circuit-board, CPU, [[(lo, 201, (10, 15,20,2211, 1(20,25), (20,25,25,30)3, [(25, loo), (5 , 15, 15, 15)Il). where the units for frequency and level are assumed to be MHz and dBmV/m/Hz respectively. Fuzzy Variable Representation I Fig. 2. Trapezoidal membership function for fuzzy attributes Each volume node may have one or more susceptibility attributes (S) associated with it as well. For example the susceptibility attribute for a CMOS gate might be represented as: Notice that both susceptibility and disturbance attributes are represented in the form: (Attribute, Type, Sub-type, [list of fuzzy representation]). These are stored in an electromagnetic properties database which can be loaded and edited by the user if necessary. If at a future time, more precise models are derived for an attribute only the database needs to be changed since the attributes are loaded into the topology via their Type and Sub-type labels. I n a similar way each surface node will have shielding effectiveness (SE) attributes associated with it. These attributes have the same form as the susceptibility and disturbance attributes, but represent the amount of attenuation a disturbance encounters while crossing from one volume to another via that surface path. The units for this quantity depend on the two nodes which the path connects (i.e. ff-path, fc-path, cf-path or cc-path). Again, more than one attribute may be associated with a surface node. This would indicate parallel paths of entry from one volume to another. The total disturbance, susceptibility and shielding effectiveness representations for a node are derived from the fuzzy representations of all the disturbances and susceptibilities present in that volume. The individual attributes are frequency range normalized to a user specified global frequency range list and added in parallel to determine the total disturbance, total susceptibility and total shielding effectiveness for the node as shown in Fig. 3. This procedure is analogous to that described in [ l , 21 with fuzzy variables replacing fixed discrete intervals. Volume Node or Surface Node ,