The objective of this paper is to develop an efficient meth- od for the construction of the model for PLC simulation, which is able to generate a virtual plant model of the produc- tion system automatically。 The proposed method employs a reverse engineering approach to extract general log data from the existing production system, which is described at the level of sensors and actuators。 The method can reduce the time and

effort required to construct a virtual plant model for PLC simulation。 Traditionally, research on virtual engineering has focused on the forward engineering approach, which is the traditional process of moving from high-level abstractions and logical designs to physical implementation of a system [20]。 In contrast, the proposed reverse engineering approach for PLC simulation is defined as a process of obtaining physical or logical implementation data from PLC log data acquired by digitizing an existing production system。 This approach can reduce the construction time and effort required to obtain a virtual plant model of a newly planned production system that references an existing production system and uses it to analyze the process of a production system at the control (sensor and actuator) level。

The overall structure of the paper is as follows。 Section 2 presents the overall approach to the automatic generation of a virtual plant model of a production system using reverse engineering, while Section 3 details an algorithm of the pro- posed method。 Section 4 shows an example and illustrations。 Finally, concluding remarks are given in Section 5。

2A reverse engineering approach for PLC simulation

Various machines that operate simultaneously in a production line are usually controlled by PLCs, which is currently the most suitable and widely employed industrial control technol- ogy [12, 18, 19, 21]。 A PLC emulates the behavior of an electric ladder diagram。 PLCs use an input/output signal table and a scanning cycle, as they are sequential machines, to emulate the working of parallel circuits that respond instanta- neously。 When a program is run on a PLC, it continuously executes a scan cycle。 The program scan solves the Boolean

Fig。 2  An implementation procedure of a virtual plant model

are highly compatible with object-oriented specifications for simulation models。 A brief explanation is given below。 Within the DEVS formalism, an atomic model M is specified by a seven-tuple:

M ¼ 〈X ; S; Y ; δint; δext; λ; ta〉

X Input events set

S Sequential states set

Y Output events set

δint S→S: internal transition function

δext Q*X→S: external transition function

logic related to the information in the input table with that in the output and internal relay tables。 In addition, the informa-

Q={(s,e)| s ∈ S 0≤e≤ta

(s)}

Total state of M

tion in the output and internal relay tables is updated during the program scan。 In a PLC, this Boolean logic is typically represented using a graphical language, known as a ladder diagram [22]。

As depicted above, to implement a VC-based PLC simu- lation model, it is necessary to construct a virtual plant model for all behaviors of a real production system。 In a conventional implementation procedure of the virtual plant model, the analyzing and modeling phases are performed sequentially, as shown in Fig。 2。 In the analyzing phase, a set of tasks that are assigned to the device can be identified, after which a pair of PLC I/O signals (input and output) can be allocated for each task。 The activation of each task is normally triggered by an external signal from the PLC program。 Once both the set of tasks and PLC I/O signals are defined for the device, it is possible to extract the state transition diagram, which defines an atomic model of the discrete event systems specifications (DEVS) formalism [21, 23]。 The semantics of the formalism