Introduction

The present invention is a method for testing a control system. More specifically the invention is an improved method for real-time testing of a control system

controlling a complex real world system.

Background

Control systems typically comprise a combination of hardware and software for controlling a physical system also called a real world system. The hardware comprises means for interfacing to the real world system that it is controlling, and the software is for managing the operations by generating signals for controlling the real world system via the interface.

Testing of the control system can be performed by simulation software simulating the real world system. Bugs in the controller software can then be detected and corrected. Testing also involves ensuring that the software gives the desired functionality to the user.

Conventional ways of testing controllers are static testing where the controller is tested separately from the real world system that it is designed to control. Such techniques can however not provide the same results and efficiency as real-time testing of control systems. Real-time means the actual elapsed time.

Real-time software systems have strict timing constraints and have a deterministic behavior. This is because real-time software systems have to schedule their tasks such that the timing constraints imposed on them are met. A conventional static way of testing is not adequate for dealing with such timing constraints. Real-time testing is thus important.

Real-time testing of control systems is a technology well explored in industries such as automotive or aerospace. It has recently also been applied in the offshore drilling industry which is driven by increased complexity of machinery, strict

commissioning time requirements and the level of severity in case of failures.

A real-time test system can be characterized as a simulated process which replaces the controlled process either fully or partially, and which is operated with real control hardware such as for instance a Programmable Logic Controller (PLC). Real-time testing systems are also known as real-time hardware in the loop (HIL) testing. The concept is illustrated in Fig. 1 illustrating that the control system is connected to either the real world machine or the model simulating the machine.

Real-time simulation testing means that a virtual model of a real world system is simulated at the same time-rate as the actual real world machine. That is the simulation time used by a computer for executing a simulation of a process performed by a system is the same as the real-time elapsed for the process.

Prior art methods for real-time testing of control systems are characterized by simulations of machinery and processes realized in an environment only close to real-time, or models of offshore machinery which are relatively simple and that are easily simulated on a typical computer.

There is a need for a method for realistic real-time testing of complex control systems. In order to perform real-time testing it is vital to assess if the simulation time is equal to the time elapsed by monitoring real-time performance of a simulator. The present invention is an improved method for ensuring real-time performance of a simulation for testing a control system connected to a complex real world system.

Short description of the invention

The present invention is defined by a method for ensuring real-time testing of a control system or at least one part of a control system for controlling a process or at least a part of a process of a real world system.

The control system or the at least a part of the control system is connected to a simulator for simulating the real world system.

The real world system is simulated in real-time by applying a modified variable step-size solver.

Real-time performance of the simulator is monitored, and step-sizes used for the modified variable step-size solver for running the simulation are adapted by:

- comparing the time used by the solver for simulating the real world system with a real-time clock;

- setting the maximum step- size of the solver based on the sample rate of the control system, and

- forcing the solver to return a simulation result when the control system demands it. Other features of the invention are described in the dependent claims. Detailed description of the invention

The invention will now be described in detail with reference to the figures where:

Figure 1 illustrates the set-up for real-time testing, and Figure 2 illustrates the components of a real-time simulator.

Figure 1 illustrates the set-up of the elements for real-time testing where the control system can be connected to either the real world machine or the model simulating the machine.

The controller will send the same controlling signals to a real world machine to be controlled or a simulator simulating the real world machine. The control system will thus not be able to differentiate between the two.

Figure 2 illustrates the two main components of a real-time simulator and their connections to the real control system controlling a real world system. The simulator is a hardware device such as for instance a computer running modeling software with all modules necessary for simulation a specific real world system.

The invention comprises a method for ensuring real-time testing of a control system connected to a complex real world system for controlling this. A complete control system for controlling a real world system can be tested, or at least a part of a control system for controlling at least a part of a process of a real world system. A process to be controlled includes a continuous, discrete, and batch process.

The control system for controlling said real world system is connected to a simulator for simulating the real world system. They are either connected via an interface or directly connected to each other. The control system to be tested is the same as the one controlling the real world system. In one embodiment of the invention, the real world system is offshore drilling equipment. Other types of systems are also feasible.

In one embodiment of the invention, the real-time simulator is running on a computer. In another embodiment, the real-time simulator is running on dedicated testing hardware. In yet another embodiment of the invention, the real-time simulator is running on a controller other than the real-time controller that is tested.

When the control system for controlling said real world system, or at least a part of it, is connected to a simulator for simulating the real world system, or at least a process of the real world system, the real world system is simulated in the simulator in real-time by applying a modified variable step-size solver that will be explained in the following. The type of solver used is the main factor which decides if a large simulation model can be utilized in a real-time test. In order to solve a given mathematical problem comprised in a simulation model, it splits the whole simulation into small portions, so called step-sizes. In principle, there are two types of solvers where one is the fixed step- size solver and the other is the variable step- size solver.

The fixed step-size solver keeps constant time steps during code execution. This is however problematic if a high level of accuracy for a simulation is required or if a large model is to be simulated.

The variable step-size solver, on the other hand, adjusts the step-size throughout the calculation to ensure that both satisfactory level of accuracy and increased computational efficiency are achieved.

The present invention comprises a method for combining the benefits of the two solvers and developing what is called a modified variable step-size solver which allows for accurate simulation of large models and simultaneously enables real-time execution.

In order to verify real-time performance of a simulation, the elapsed simulation time used by the solver for simulating the real world system and the time are measured and compared by subtraction. A convenient factor for measuring real-time performance of an application is the total delay, i.e. the difference between the real- time and the simulation time. For a real-time application it is expected that this value is constant and close to zero throughout a simulation.

As said the real world system is simulated in real-time by applying a modified variable step-size solver. While the simulator is running, real-time performance of the simulator is monitored and step-sizes used for the modified variable step-size solver for running the simulation are continuously adjusted and adapted.

As said the time used by the solver for simulating the real world system with a realtime clock is compared. If subtracting the simulation time from the elapsed time yields a positive value (e.g. computing 1 minute of simulation takes 10 minutes) the simulated system is too complex and too slow to be used in a real-time hardware in the loop (HIL) test. This is a typical situation when a detailed model of a system applied in offline simulations is directly exported to a real-time simulator. This setting is problematic if a high level of accuracy is required or if a large model is to be simulated. In addition, this normally occurs if a fixed step- size solver is used. If a simulator lags behind the real-time it cannot be used in real-time HIL tests. A solution is to simplify a complex model and make it run in real-time with an acceptable level of accuracy. Simplifying a model is however not always possible or desirable. If on the other hand subtracting the simulation time from the elapsed time yields a negative value (e.g. computing 10 minutes of simulation takes only 1 minute) the simulated system is too fast and can not be used in a real-time HIL test. Even if the obtained results are accurate such a model is not suitable for real-time simulation due to the time stamps discrepancies. However, since it is confirmed that such a simulation yields accurate enough results, the concern is its real-time performance which could be achieved by limiting the maximum step- size of a solver. The parameters and step-sizes will depend on model complexity and the desired level of accuracy. This situation is less problematic than the example above. If a solver has the results of a simulation ready faster than the real-time it will wait for the next cycle to begin. This is achieved by setting the maximum step-size of the modified variable step-size solver to a value which will prevent the solver from going faster than the real-time.

Decreasing step sizes means running more parts of a simulation model giving more accurate results, and increasing step sizes means running less parts of a simulation model giving more coarse results.

The real-time clock of the real world system will define an upper limit for maximum time the solver can use. This upper limit will provide an input to the simulator running the modified variable step size solver. According to the present invention, the maximum step size of the solver is also set, and where this is based on the sample rate of the controller.

The last step of the inventive method is to force the solver to return the simulation result and when the controller demands this based on the elapsed real-time. The solver can also return the corresponding simulation accuracy level. The accuracy levels of the simulation result are calculated using local truncation error techniques which are commonly used in solver algorithms.

If the simulation accuracy levels are considered to be too low the simulation model may be too complex for real-time testing.

The inventive method comprising the different steps defined above will ensure that the simulation will be run in real-time and the accuracy of the simulation result will be presented.

Using a modified variable step- size solver according to the invention realizes accurate simulations of large models running in real-time.

The inventive method is a significant step forward in the field of real-time testing, and is well suited for real-time testing of control systems for real world hydraulic- mechanical systems used on for instance drilling rigs. Simulations of machinery involving real control systems have in this area so far been realized in an

environment only close to real-time or with models of offshore machinery that are relatively simple and easily simulated.

The inventive method is also well suited for real-time testing of a control system for a real world electrical system or for real-time testing the control system for a real world system comprising a combination of mechanical and electrical systems.

By using the inventive method there is no need to explicitly formulate a detailed set of mathematical equations describing the system to be simulated. Instead, the model can be created directly in a virtual simulation environment by using standard multi- domain libraries and components. The solver can then use libraries and modules which emulate real world behavior of the real world system.

Such a testing framework will offer improvements over existing solutions in terms of simulation fidelity and accuracy levels as well as modeling efficiency and complexity. The inventive method can be successfully applied for testing and examining real control systems of large machines in a virtual simulation environment, and has good scalability for different hardware platforms.