Preface Modern industry faces increasing challenges: fierce market competition and stringent government regulations on environmental protection. The main strategies for addressing these challenges are reducing new product design and testing time, minimizing development costs, and shortening product launch cycles. The key to solving these problems lies in the effective utilization of CAD/CAE/CAM technologies in product development and process design. Among the two main problems that most perplex designers and developers of electromechanical products, the first is how to rationally select components and subsystems to meet system requirements, and the second is how to test system performance under low-cost conditions. Simulation technology has emerged as the answer to these problems. In the past, research on simulation technology mainly focused on simulation tools, while the accuracy and functional precision of simulation models were relatively neglected for modern electromechanical systems. No matter how advanced the simulation tools are, an unsuitable simulation model will inevitably lead to erroneous simulation results. This is a major problem facing modern simulation technology. Virtual instrumentation technology uses mainstream computer technology and combines innovative, flexible software modules with high-performance hardware technology to create powerful computer-based instrumentation solutions. NI has released a complete set of software and hardware tools for building measurement and control applications. They provide a solid foundation for the development of simulation systems and models. The research on runtime environment simulation systems initially focuses on simulation models, precisely establishing a dynamic case simulation model library for electromechanical components and systems. This dynamic simulation model library has three advantages: a) it provides a simulation model classification structure based on case-based reasoning technology, allowing for dynamic adjustment of the model database structure to meet actual system requirements; b) it has the function of testing the matching degree between simulation models and real systems; c) it has a system identification tool that can extract accurate models from actual experimental results and reconstruct the dynamic model library. These advantages effectively solve the problems previously caused by static model libraries in simulation experiments: the dynamic simulation model library has dynamic and adaptive capabilities, and can meet a wider range of needs. Another focus of the runtime environment simulation system is the establishment of a hardware-in-the-loop test system. Based on virtual instrument technology, this work becomes easier. The specific structure and development methods will be introduced later.
**[b]Structure and Development Method of the Runtime Environment Simulation System[/b]** The runtime environment simulation system comprises two parts: a dynamic case model library and a hardware-in-the-loop testing system. The specific structure is shown in Figure 1. The dynamic case model library mainly includes several modules: an index engine, a mathematical model library, a data description library, other databases, a dynamic identification module, and a model testing and evaluation module. The hardware-in-the-loop testing system can be divided into software and hardware components. The software component mainly consists of measurement and control software modules; the hardware component includes all hardware systems, such as PXI bus systems, PCI bus systems, and Compact-RIO systems.
Dynamic Case Model Library The development of the dynamic case model library is primarily based on the NI LabVIEW simulation module, Matlab, and other simulation toolkits. These tools provide a complete platform for building a model library for electromechanical systems. The dynamic case model library consists of three main parts: the case model library (index engine, mathematical model library, data description library, and other databases), the system identification software module, the model testing and evaluation module, and the model transfer and modification module. The main structure is shown in Figure 2.
Case Model Library The model library is implemented using case-based reasoning technology, which gradually gained attention from more and more artificial intelligence researchers in the late 1980s. It is an analogical method for solving problems using past cases and experience. In general, case-based reasoning technology adopts the following reasoning steps: identifying the problem, acquiring cases, modifying cases, and storing cases. The most important part of case-based reasoning technology is building an index engine for cases and designing an indexing algorithm. We can use this technology to build a case database. The entire database can be built into a distributed network structure with reconfigurable characteristics. Its main advantage is that it can reconfigure the distributed network according to user needs and quickly guide users to suitable cases. This technology has strong adaptability. In this model library, the models of electromechanical systems and components can be divided into several main parts, such as mechanical, electronic and electrical, and hydraulic. These categories are assembled into a tree structure. For example, electronic and electrical components can be divided into microprocessors, actuators, drive systems, sensors, etc. Furthermore, actuators can be further classified according to different types, power, maximum speed, drive methods, etc. Figure 3 shows a simplified example of motor classification. Case 1 represents an original model in the model library. Based on the above classification, we can obtain various fast and effective indexing methods for different models. For example, as shown in Figure 3, there are currently three cases in this model library, each representing a type of motor model. If we now need to obtain an AC motor model with a power greater than 1kW and a maximum speed greater than 3000rpm, but there is currently no matching model in the library shown in Figure 3, the case model library will automatically create a new case and reconstruct the model library. The reconstructed model library structure is shown in Figure 4. The above example explains the method of model library reconstruction. Model Testing and Evaluation Module The model testing and evaluation module mainly includes two parts (see Figure 2). The first part compares the simulation experiment results with the actual experiment results and determines the differences between the two; the second part tests and evaluates whether the simulation model corresponds to the actual components and system. After obtaining the experimental results, the case model database will automatically select relevant models and simulation experiment results to provide to the testing process. The two functions of model testing and evaluation are described as follows: Comparison function: An effective way to compare simulation experiment results and real experiment results is to calculate the difference between the output data of the two. We can also compare the errors in system parameters, performance indicators, dynamic feature maps, etc. After comparison, these errors are provided to the model testing and evaluation module. The model testing and evaluation function determines whether the error sequence is a white noise sequence with zero mean and very small variance. If so, the simulation model is very close to the actual system, and the model does not need modification or reconstruction. Otherwise, the simulation model must be modified or reconstructed to improve matching accuracy. The dynamic identification module then takes over if the model testing and evaluation results indicate that the relevant model must be modified or reconstructed. This module uses real experimental results to obtain a new model and performs model calibration and modification. Modern system identification theory plays a crucial role in this module. System identification mainly obtains an equivalent system (mathematical model) based on the input and output of the system being identified. Common model description methods include transfer functions, state equations, and differential equations. Transfer function identification methods are divided into time-domain and frequency-domain methods. State equation identification methods are more complex and can be derived from differential equations or transfer functions. Differential equation identification mainly employs statistical analysis and parameter prediction methods, such as least squares and maximum likelihood estimation. Nonlinear systems can be described using nonlinear differential equations, Volterra series, bilinear models, and other methods. Different models are needed for different components and systems. Even for the same component, different description methods are required to meet different needs. In the initial stage of system identification, the first step is to select the correct and appropriate mathematical model type based on actual requirements. Then, the next step is to select a suitable identification method to obtain model parameters from actual experimental data. For example, a motor can be described as a linear model or a nonlinear model. Depending on the required simulation accuracy and functionality, we can choose a single-input, single-output transfer function or use the least squares method to construct a Volterra series model. Many modern artificial intelligence theories can be used for identification here, such as artificial neural networks, fuzzy logic, H∞, and genetic algorithms.
The system identification module primarily performs the following functions: Selecting a suitable mathematical model to describe the actual system based on the required simulation accuracy and simulator functionality; Selecting a suitable system identification method to obtain the required model parameters and other descriptions; Testing the accuracy of the mathematical model. After the above identification work, the improved or new mathematical model will be added to the case model library. The work of the case model library, model testing and evaluation module, and dynamic identification module forms a closed loop, thus ensuring the adaptive performance of the entire model library and constituting the entire dynamic case model library. Hardware-in-the-Loop Test System The hardware-in-the-loop test system was initially conceived as a single-function test system, primarily used in the automotive industry for testing engine control components. Now, more and more electronic control components and other general testing applications are gradually adopting hardware-in-the-loop technology. The main problem facing engineers in establishing hardware-in-the-loop test systems is how to synchronize the simulation system and the actual system through a large number of high-speed I/O channels and signal conditioning channels, while ensuring functionality and performance. With the increasing functionality and flexibility of computers, engineers and researchers are increasingly inclined to use virtual instruments to implement hardware-in-the-loop test systems. [align=center]The implementation of hardware-in-the-loop (HIL) test systems is becoming increasingly easier through the use of virtual instrumentation technology. Figure 5 shows a concrete example of a HIL test implementation. This program, implemented using LabVIEW, embeds the testing process into a mathematical model described by transfer functions, thus realizing a HIL test flow that blends the actual system and the mathematical model.[/align]
This implementation method allows many electromechanical products to be introduced into the measurement and control system through different description methods, including mathematical models, data tables, data graphs, etc. Through organic integration with a dynamic case simulation model library, these models will be added to the simulation model library. In this way, the hardware-in-the-loop testing system has a robust model library as a resource center. Introduction to the Washing Machine Main Control Board Test System The main control board is the core control unit of a fully automatic washing machine. After assembly, the functions and input/output interfaces of the main control board must be tested. The main test targets include door opening/closing signals, shut-off signals, water level signals, inlet valve control signals, drive control signals, motor control signals, etc. To accomplish the above tasks, the test system needs to generate simulation signals itself. Therefore, we developed a washing machine main control board test system based on a runtime environment simulation system. This system uses NI multifunction boards and the LabVIEW software platform to simulate the normal working state of the washing machine and perform fully automatic testing. Below, we will introduce the development method of this representative runtime environment simulation system. First, we extract simulation models of components closely related to the main control board, such as frequency-discriminating water level sensors, motors, and inlet valves. Among them, the frequency-discriminate water level sensor is very representative. Below is a brief introduction on how to establish a simulation model of the frequency-discriminate water level sensor. The structure of the frequency-discriminate water level sensor is shown in Figure 6. It uses an LC electromagnetic resonant circuit as the sensing element, converting the water level signal into changes in LC parameters, and finally outputting a frequency signal. The principle can be simply described as follows: the water level first affects the air pressure in the air chamber. The change in air pressure causes the guide plate to move, and the magnetic core moves within the coil, which changes the inductance of the coil. Ultimately, the LC circuit generates different frequency signals. The equivalent circuit of the frequency-discriminate water level sensor is shown in Figure 7. Generally, after the frequency-discriminate water level sensor is installed and fixed, its coil turns, air permeability, magnetic core permeability, average coil radius, effective magnetic core radius, and coil length remain unchanged. The only change is the position of the magnetic core within the coil. This movement is linear, resulting in a continuous change in inductance. Through theoretical analysis and experimental verification, the water level signal and the output frequency of the frequency-discriminate water level sensor are inversely proportional. Below are characteristic tables for two commonly used frequency-discriminating water level sensors, describing the correspondence between water level and output frequency. The first is the SW-4 type, and the second is the XQB52-108G type. Through experimental results and mathematical analysis, we obtained mathematical description methods for different frequency-discriminating water level sensors. This method was also applied to the simulation mathematical description of other components, and these mathematical descriptions were added to the dynamic case simulation model library for the next step of building a fully automatic washing machine main control board test system. Subsequently, we developed a complete hardware-in-the-loop test system using virtual instrument technology. NI's M-series multifunction cards were selected as the data acquisition module, and the entire test software was developed using NI's LabVIEW software platform. Based on the hardware output, the signals of the main sensors and actuators, such as door opening/closing signals, shut-off signals, water level signals, inlet valve control signals, driver control signals, and motor control signals, were simulated. Finally, the fully automatic testing of the washing machine's rinsing, washing, and spin-drying states was completed. Conclusion The operating environment simulation system has two important components: a dynamic case simulation model library and a hardware-in-the-loop test system based on virtual instruments. The former primarily expands the mathematical description scope of real components and systems, employing case-based reasoning for reasonable classification; the latter focuses on how to implement practical measurement and control systems using mathematical description methods to reduce development difficulty, costs, and other investments. These two parts are closely related, but they are relatively independent systems. The runtime environment simulation system bridges this gap, allowing real-world objects to be closely integrated with simulated objects in the virtual runtime environment, forming a higher-level measurement and control system. Future work will focus on the two aspects described below: first, improving the quality of model building and indexing engines, expanding the types of simulation models, and simplifying mathematical description forms; second, further improving hardware-in-the-loop simulation technology, real-time measurement and control environments, and distributed communication technologies. These two aspects will undoubtedly elevate runtime environment simulation technology to a higher technical level, providing better and more practical development and testing tools for modern industry. Editor: He Shiping
