pc_ieee_trans_consumer_elect_1995.
Contents
1. Corcoran P M and Lusted K A Remote Electronic Object Emulation System for Home Bus Applications IEEE Trans Consumer Electronics Vol 40 Aug 1994 pp 405 410 8 Forgy C L OPSS User s Manual Technical Report CMU CS 81 135 Department of Computer Science Carnegie Mellon University Lusted et al An Active Node Fault Diagnosis System for CEBus Networks Biographies Peter Corcoran received the BAI Electronic Engi neering and BA Maths degrees from Trinity Col lege Dublin in 1984 He continued his studies at TCD and was awarded a Ph D in Electronic Engineering in 1987 for research work in the field of Dielectric Liq uids In 1986 he was appointed to a lectureship in Electronic Engineering at UCG His research interests include microprocessor applications environmental monitoring technologies and automated testing of electronic components and equipment He is a mem ber of the IEEE Karl Lusted received his BE degree in Electronics from University College Galway in 1992 He is pres ently studying for the degree of M Eng Sc at UCG His research interests include microcontroller applica tions and communications network protocols He is a member of I E I and an associate member of I E E 889 Paul J Nolan is a Statutory Lecturer in Mechanical Engineering at University College Galway He re ceived the BE degree from University College Dublin and the M Eng and PhD degrees in electrical en2. identifier for the event and various attributes of the event such as time event type packet contents etc rec no 570 date 28 05 95 time 21 43 00 source 0 destin 0 ust del 0 mess P type 00 prev 569 plist empty rec no 571 date 28 05 95 time 21 43 02 source 3 destin 2 ust del 0 mess P type 09 prev 570 plist 70 E7 OF F4 30 34 F5 FF FF FF FF rec no 572 date 28 05 95 time 21 43 02 source 0 destin 0 ust del 0 mess P type 00 prev 571 plist empty Figure 5 Sample data with WME elements Rules in the expert system comprise one or more condition elements CEs and actions The CE s are the IF part of the rule and the actions are the THEN parts The syntax is quite straightforward and is de scribed in 8 It should be strongly emphasised that the results pre sented so far could have been implemented algo rithmically e g using a BASIC C program or using the search facility in an available data base The power of an expert system becomes clear when the rules for the underlying CEBus objects and the appro IEEE Transactions on Consumer Electronics Vol 41 No 3 AUGUST 1995 priate CAL instructions are added to the rule base The inclusion of heuristics information and the intelli gent application of the active mode also requires the expert system The diagnosis implemented to date has used rules explicitly specified by engineering knowl
3. NUI Galway O Gaillimh ARAN Access to Research at NUI Galway Provided by the author s and NUI Galway in accordance with publisher policies Please cite the published version when available An Active Node Fault Diagnosis System for CEBus Networks Corcoran Peter Nolan Paul J Humborg Kenn Lusted Karl Publication Date Lusted K Humborg K Nolan P J amp Corcoran P M Publication 1995 An active node fault diagnosis system for CEBus Information networks IEEE Transactions on Consumer Electronics Vol 41 No 3 pp 884 889 Publisher http hdl handle net 10379 275 Some rights reserved For more information please see the item record link above Downloaded 2015 12 17123 38 452 Manuscript received June 12 1995 884 IEEE Transactions on Consumer Electronics Vol 41 No 3 AUGUST 1995 AN ACTIVE NODE FAULT DIAGNOSIS SYSTEM FOR CEBus NETWORKS Karl Lusted Kenn Humborg Paul J Nolan and Peter M Corcoran Dept of Electronic Engineering University College Galway Dept of Mechanical Engineering University College Galway Abstract The design and implementation of an active node fault diagno sis system for CEBus networks is described The system uses an IBM PC with a user friendly GUI to provide statistical infor mation on network traffic and messages The PC is linked to an active CEBus node which monitors the network and can initiate testing sequences controlled b
4. d A de scription is then given of the fault diagnosis system This consists of a CEBus node with a physical inter face to the AC power line 2 This node is controlled by a dedicated microcontroller with modified protocol software to facilitate test and monitoring activities This active node is linked to an IBM PC which con trols the analysis of network monitoring activities and presents data summaries and control functionality to the system User via a GUI interface The fault diag nosis system overview is given in Figure 1 E Fauit Diagnosis le Software K i y Eoy y o y Diagnostic Routines Graphical User Test Sequences etc Interface Database l i x D Va i ERA A _______ _ Message Transfer Element A yY eo CEBus Nod Modified protocols see text AC Powerline Figure 1 CEBus Fault Diagnosis System Overview 2 CEBus Packet and Data Formats This section describes the communications interface between the modified CEBus modem and the PC used for the actual diagnosis of network fault conditions This is important for two reasons firstly some error detection is carried out in the hardware implementing the physical layer of the CEBus interface and this in formation can be trapped and passed on to the PC software by the CEBus active node and secondly the packet formats described in this section provide our second lev
5. dically and random packet generation is also supported The remaining two nodes were connected to IBM compatible PCs and controlled by software running on the PCs One node was used to implement the fault diagnosis system The other node was used to simu late faults by transmitting invalid packets and disobey ing the channel access protocols 4 2 Active Node Software The microcontroller subsystem implements the data link and network layers of the CEBus protocols For the present work they were modified to facilitate the requirements of the active and passive modes of the fault diagnosis system The modifications allow the node to capture packets from all other nodes on the system when it is in passive mode In active mode it can simulate another system node Test packets can also be generated by the node to allow active testing of the network This feature allows periodic or continuous testing of a suspect node on the network and would be particularly useful for the detection of intermittent fault conditions A graphical user interface GUI was developed for the PC software This gives the user access to all the available functionality of the system in an easy to learn and user friendly manner The main screen of this interface is shown in Figure 3 Plotting functions have also been incorporated into the software to pro duce graphs such as Figure 4 09 03 01 02 OL 70 E OF F4 30 34 F FF FF FF FF FD 7D C Carrier fost w
6. edge of the CEBus protocol and the likely faults An alternative approach which may be necessary is to use rule in duction from a large example set This will be used to complement the rules deduced above 7 Conclusions A working CEBus fault diagnosis system for the AC powerline medium has been developed The system has been designed to monitor network traffic and per form a range of analyses for actual and incipient fault conditions It provides a useful aid to the developers of CEBus devices and has the potential to be applied as a practical tool in the debugging of faulty CEBus networks REFERENCES 1 Hofmann J The Consumer Electronic Bus An Inte grated Multi Media LAN for the Home International Journal of Digital and Analog Communication Sys tems Vol 4 No 2 Apr 1991 pp 77 86 1991 2 EIA Home Automation System CEBus Interim Stan dard IS 60 Vol 1 Parts 1 7 8 EIA June 29 1992 3 EIA Home Automation System CEBus Interim Stan dard IS 60 Vol 4 EIA June 29 1992 4 Pakkam S R and Manikopoulos C N Performance Evaluation of the Consumer Electronic Bus EEE Trans Consumer Electronics Vol 36 Nov 1990 pp 949 953 5 Yang J and Manikopoulos C N Performance Com parisons of the CEBus with other Protocols JEEE Trans Consumer Electronics Vol 39 Nov 1993 pp 824 831 6 EIA Home Automation System CEBus Interim Stan dard IS 60 Vol 8 EIA June 29 1992 7
7. el of error check on the transmission integ rity of the CEBus network 0098 3063 95 04 00 1995 IEEE 2 1 CEBus Modem to PC communications There are two principle forms of data transmitted from the modem to the PC i Single characters indicating errors detected by the CEBus modem These errors are generally physical layer errors or errors in node operation e g packet collision on network communications buffer full etc ii Received packets detected on the AC power line medium Received packets consist of standard CEBus DLL frames received over the network en capsulated with Start_of Packet and End of Packet characters These extra characters serve to distinguish the CEBus packets from the character based communication of detected errors The standard CEBus frame format using SE error detection is given in Figure 2 PRE Packet preamble Control Control field packet type priority service class DA Destination Address DHC Destination House Code SA Source Address SHC Source House Code Information Information field Figure 2 Normal MAC Frame Format 2 2 PC to CEBus Modem Communications In active mode the fault diagnosis system transmits test packets over the AC powerline These test packets are designed to test for a particular type of error sug gested by analysis of received packets during passive mode The PC sends DLL frames preceded by a Net work Layer header and any requests for Network Laye
8. gineering from McMaster University in Ontario His employment included work at Power Technolo gies Inc Schenectady NY His research interests include computer simulation and control and applica tion of AI in engineering Kenn Humborg received the BE degree in Electronic Engineering from University College Galway in 1994 He is presently studying for an M Eng Sc de gree at UCG His research interests include computer networking and application of fault diagnosis methods in automatic test equipment
9. hile RXing from mains ac 00 40 00 80 00 00 06 OC CO 80 48 2C 6a FD OL OL 03 OL 70 OF F4 30 30 FS FD oo FD 09 02 01 0l OL 70 E OF F4 30 30 FS FD 09 02 0l OL O1 70 E OF Fa 30 30 FS FD FD 09 03 01 02 OL 70 E OF F4 30 34 FS FF FF FF FF FD FD 09 03 01 02 OL 70 EF OF F 30 34 FS FF FF FF FF FD 28 05 95 21 41 12 09 03 0l O2 OL 70 E OF F4 30 34 FS FF FF FF OFF FD Figure 3 The User Interface Channel Activity trom 28 05 95 21 40 23 10 26 05 95 21 45 15 Packets Events sec 1 Errors p LAA AMILA AMMA M AL Time Figure 4 Graph of network activity Lusted et al An Active Node Fault Diagnosis System for CEBus Networks 5 Fault Diagnosis System Operation Initially the system simply listens in on the CEBus network recording all network events packets and physical layer errors and signals and their associated timings and storing this information to a database The system then analyses the collected data in an attempt to detect faults or determine probable fault scenarios The analysis engine produces a set of packet sequ ences designed to confirm or disprove its hypotheses and to collect additional data if required After the analysis stage the system transmits the test packets and records network events The analysis stage is performed again to give an improved diagno sis The process can be repeated as often as desired The test sequences of packets transmitted over the network are re config
10. ies on a set of grammatical rules for CAL It is also possible for devices to send CAL messages that refer to objects on a device that do not exist refer to Vs within an object which do not exist etc The standard specifies that each node can be asked to supply details of its constituent objects The system can thus build up an internal representation of the devices of the network 7 The validity of CAL mes sages can be checked by using this information in the analysis stage 4 The Active CEBus Node The active CEBus node is controlled by a dedicated microcontroller with modified protocol software to facilitate test and monitoring activities The Medium Access Control MAC sublayer of the data link layer has been modified to allow the system to receive and record all network packets These captured packets are passed to the Application layer along with channel timings in Unit System Times USTs Network errors detectable by the lower layers of software in the CEBus node are also passed to the Application layer for storage to an integrated database 4 1 Active Node Hardware This consists of three main functional subsystems interface circuitry to the mains powerline a spread spectrum mains modem IC and a supervisory micro controller A small CEBus network was implemented using five CEBus mains modems Three of the modems can be operated in standalone mode In this mode packet sequences can be programmed to repeat perio
11. iming but also the packet type to see if the Channel Access Method is being adhered to The Channel Access Method also specifies additional de lays for packet priority packet queuing and randomi sation By examining the inter packet timings and the packet data containing priority information etc the fault diagnosis system can detect more subtle errors in the Channel Access Method 3 2 2 Failure to send ACK packet If a CEBus node transmits a packet and the packet data indicates that an ACK packet is expected then the next packet received should be an ACK packet from the receiving node The fault diagnosis software must examine all packets to see if an ACK packet is requested and then check the next packet to see if an ACK packet was sent Failure to do so suggests that the receiving node is faulty and can be checked in active mode by the transmission of packets to the sus pected faulty node 3 2 3 Multiple transmission of the same packet If a CEBus node continually transmits the same packet to another CEBus node even though the origi nal packet was received correctly the transmitting node may be faulty This error can only be detected by creating a database of all original packets on the CEBus network during the passive mode analysis A new packet is added to this database only if it can not be found in the database already Each time a match is found for a particular packet a counter for that packet is incremented Thus the e
12. lar expert system used in this project is pattern driven in that it comprises e working memory 887 888 e production memory rules e an inference engine The expert system thus works by trying out rules based on the current state of the working memory using the inference engine and updating the working memory as rules are fired This approach is most suit able for general monitoring and analysis of the CEBus traffic In particular it is expected to be particularly appropriate in analysing the application layer where there will be a very large number possible solutions involving various objects instance variables IV s and CAL language constructs The alternative ap proach diagnosis or backward search approach can also be readily implemented In this case the goal is placed in working memory and matched against the actions of the production rules This strategy is gen erally more suitable for diagnosis applications where there 1s a smal number of possible outcomes The generality of system in implementing both forward and backward search or a combination of both and well as the ease with which heuristics can be included are important at this stage of development of the proj ect The data collected by the system is converted into a form suitable for input into the expert system Each event is represented as a Working Memory Element WME A sample of the WME file is shown in Figure 5 Each WME comprises a unique
13. r services from the Application Layer diagnosis system 3 Classification of CEBus Network Errors It is necessary to classify CEBus network errors in order to develop a fault diagnosis system to detect them CEBus errors have been grouped according to protocol layer the list is not complete but does give an indication of the type of errors that can be expected to occur Lusted et al An Active Node Fault Diagnosis System for CEBus Networks 3 1 Errors Generated in the Physical Layer The Physical Layer is the lowest layer of the CEBus protocols It is concerned with the physical transmis sion and reception of bits on the CEBus medium i e the AC powerline Errors occurring in the physical layer are not normally reported to higher CEBus lay ers but the CEBus modem software has been modi fied to indicate the occurrence of such errors for the purposes of fault diagnosis Elements of Layer System Management 2 techniques are used to pass informa tion from the lower protocol layers to the Application Layer Physical layer errors are useful for collecting network statistics which can provide valuable infor mation to the failure analysis software Typical Physi cal Layer errors include packet collisions loss of car rier transmission aborted etc 3 2 Errors in the Data Link Layer The Data Link Layer is responsible for the correct transmission and reception of CEBus packets The Data Link Layer handles acknowledge packets re t
14. ransmission of packets ensuring correct Channel Access Method etc Errors occurring in the Data Link Layer are usually not as easy to detect as errors oc curring in the Physical Layer 3 2 1 Incorrect Channel Access Method All CEBus nodes must adhere to a set of rules for ob taining access to the Powerline medium These rules known as the Channel Access Method stipulate cer tain delay times before attempting channel access after the last communication deferral to nodes already communicating on the network allowing time for ACK packets before attempting new communications etc Thus most of these errors can be detected by ex amining the inter packet timings sent from the CEBus Modem to PC These inter packet timings consist of the number of USTs Unit Standard Times elapsed since the last CEBus channel activity and precede the packet information Detectable errors include i Insufficient delay time between packet trans missions Generally if inter packet timing is less than 10 USTs then the communicating node is not waiting long enough between communications 3 ii Excessive delay before transmitting ACK packet Any ACK packet should be transmitted not more than 6 USTs 3 after reception of a packet requiring an acknowledgement This is to ensure that ACK packets are always sent before standard packet communications begin again 885 886 The fault diagnosis system must not only examine the inter packet t
15. urable and extendible as further network faults are categorised and their test conditions identified At present data for active test packet se quences is generated in two ways 1 Pre defined packet sequences can be defined in a data file which is loaded when the PC software Starts up 1i Additional sequences can be manually gener ated while the program is running and saved to a data file for future use Effectively the user is performing part of the analysis function It is envisaged however that the procedure will be automated to a greater degree Improved in telligence in the diagnosis engine would allow it to generate its own test sequences from a set of primi tives in much the same way as a human would Using an expert system to implement the analysis engine provides a straightforward method of encoding the necessary rules and heuristics for this type of analysis As our studies are still at a formative stage it is not yet clear how such a system could be best implemented to provide a practical balance of flexibility transparency and extendibility This must await more detailed studies on a range of CEBus networks ideally cover ing the different physical media described in the CEBus standard 2 6 Methodologies for Fault Diagnosis A basic analysis engine was developed using a readily available expert system In general expert systems are classified as either data driven pattern directed or goal driven The particu
16. xcessive transmission of a particular packet can be detected 3 3 Errors in the Network Layer The Network layer is primarily concerned with packet routing in a CEBus network which may consist of several different media linked through a series of Routers and Brouters The fault diagnosis system out lined in this paper deals with errors on a single me dium but may be extended to detect errors in the Network layer 3 4 Errors in the Application Layer CAL Error detection in the Application Layer has not been implemented in the current version of the fault diag nosis software This is principally because of the po tential complexity of CAL constructs However some preliminary studies have been undertaken and the framework of an interpretation and diagnosis system for CAL is described briefly IEEE Transactions on Consumer Electronics Vol 41 No 3 AUGUST 1995 The Common Application Language CAL provides a framework through which CEBus devices may communicate with and contro objects on another CEBus device 6 The CAL uses a particular syntax whereby every object on a CEBus device has a set of properties associated with it known as Instance Vari ables IVs The values of these Vs completely define the object in question and they may be read only or read write All CEBus devices can be represented by a collection of such Objects Mistakes and errors in the CAL syntax of CEBus packets can be detected by a system which rel
17. y the PC 1 Introduction The Consumer Electronic Bus CEBus is a multi media LAN standard developed for home automation applications 1 The CEBus standard is a subset of the OSI seven layer model In particular it conforms with the OSI model at the three lower protocol levels physical link and network layers 2 The implemen tation of these lower layers is quite robust and a prac tical CEBus network is fault tolerant with a low prob ability of communications errors propagating to the higher layers Despite this robustness communica tions errors will still impact on network performance generating increasing traffic incomplete packets and excessive usage of IACK call backs 3 In worst case situations it is conceivable that a network may fail or be rendered dysfunctional by a single faulty node Performance analysis and evaluations of the CEBus 4 5 do not allow for badly behaved nodes which can have a noticeable effect on throughput and message delay for the entire network In this paper we consider the design and implementa tion of a fault diagnosis system for practical CEBus networks The implementation described in this paper is for the mains power line medium but can be adapted for any of the media supported by the CEBus standard We begin with a summary of packet communications formats between the CEBus active node and the PC followed by an outline of the principle types of CEBus errors and how these can be detecte
Download Pdf Manuals
Related Search