Analytic Calculus of Response Time in Networked Automation Systems

Abstract

-This paper presents a novel approach to evaluate the response time in networked automation systems (NAS) that use a client/server protocol. The developments introduced are derived from modeling the entire architecture in the form of timed event graphs (TEGs), as well as from the resulting state representation in Max-Plus algebra. The various architectural stages are actually modeled in a very abstract pattern, which yields just those TEG models where local delays are sufficient to perform the overall evaluation. In this manner, linear Max-Plus equations are obtained. A thorough analysis of these equations has led to analytical formulas for direct calculus of NAS response time. As a final step, experimental measurements taken on a laboratory facility have been used to verify the validity of the results. In conclusion, the benefit and effectiveness of this novel method have been demonstrated.Note to Practitioners-In this work, we present an overall study of networked automation systems working according to client/server paradigm. Unlike systems where a global scheduling of the shared resources is available and the delays well handled, in such systems it is not the case and the investigations to evaluate their real-time performances are required. Actually, these systems are very present in industry but the efforts to deal with this issue are rare and often informal, based on simulation of particular cases. In our work, we assess a major criterion of their time performances, the response time. We give a formal evaluation of this feature through an analytic approach. The results we present are generic and fit the experimental observations in different cases.

Industrial Ethernet for Distributed Control in Factory Automation

“…However, the solutions are not completely resolved from the designer viewpoint and the applicability to industrial Ethernet networks. [20] [21] The papers [20] and [21] discuss a modeling and control system for industrial Ethernet networks, with centralized control in the master controller network, not showing or mentioning the proposal for a distributed control applied to industrial Ethernet networks. Moreover, the works [11] and [19] have already proposed, in some way, using the techniques of scheduling messages applied to Foundation Fieldbus network to perform distributed control on the bus, but did not address the treatment of faults that may occur in field modules.…”

Client-Server Networked Automation Systems Reactivity: Deterministic and Probabilistic Analysis

“…Besides that, we suppose in this paper that ratio / SCN CPU T T is integer too. Indeed, by performing the transformation exposed in the next section, we get to the same model used in [14] and therefore can consider the general case, which is already studied in [14], in a similar way. ii) Neither frame loss nor components failure is considered.…”

“…v) All the requested answers are received from the scanned RIOMs before the scanning period elapses (the network due delays are indeed often by far smaller than the scanning period) vi) According to practical observations, Proc T and T filt can be considered constant (less than 0.08% jitter [14]). …”

“…In [14], we used a model constituted exclusively of TEG to get the advantage of the resulting (Max,+) linear equations and perform an analytic study. The model of Fig.…”

“…Among the cited works, only model-checking provides both guaranteed bounds [11] and distribution [12] but unfortunately suffers from the state explosion problem. In [14], we presented an analytic approach in an effort to avoid such a problem but our investigation was limited to NAS with only one controller (we considered several remote I/O modules nevertheless). Thus, the current study is a generalization that displays two main new contributions (i) the NAS may involve multiple controllers (ii) analytic probabilistic formulae are provided to calculate the density of probability of the response using the local delays probability densities (in [14], we calculated the distribution using simulation, not analytic formulae as the case now).…”