Posted on

I. Introduction

I wanted to talk about my thoughts on the development direction of the automation industry a long time ago. If it must be traced back, it would have been more than 10 years ago, but the cautious character of technicians has always wanted to see clearly.

Point, so I procrastinated again and again, so that more than ten years passed, the procrastination was unstoppable, the eyebrows were burnt, and I decided to start the pen.

Because I am a monk with a major in electronics, I have participated in the design of the control system and taught myself halfway. I don’t have a precise control theory basis. So if there are mistakes in the article, then

It is also normal. Please correct me, especially the overall direction and ideas can be exchanged more. This will also be the direction of future development.


The automation industry is a very weird industry. The difference between theory and practice is 63 years. Why is it so accurate? Because of the classical control theory represented by Qian Xuesen’s "Engineering Cybernetics" in 1954

Generation is still the main theoretical guide in the automation industry.

Afterwards, whether it was modern control theory or the intelligent control theory proposed by some people in the 1990s, they only heard the sound, but did not see them. Most of the occasions still use PID and fuzzy control.

There are a lot of concepts, but it is difficult to promote and apply in large-scale industrial control.


Before starting our description, let's take a look at how experts define generations.


Classical control theory: a theory based on frequency method and root locus method. Classical control theory uses Laplace transform as a mathematical tool, and single-input-single-output linear time-invariant systems are the main research

Research object. Transform the differential equation or difference equation describing the system into the complex number domain to obtain the transfer function of the system, and use this as a basis to analyze and design the system in the frequency domain to determine

The structure and parameters of the controller. Usually feedback control is used to form a so-called closed-loop control system.


Features of classic control theory:


1. Classical control theory is only limited to the study of linear time-invariant systems, even the simplest nonlinear systems cannot be handled;


2. Classical control theory is limited to the analysis and design of single-variable systems, single-input single-output variable systems, which essentially ignores the inherent characteristics of the system structure, and cannot handle input and output.

Systems that are all greater than 1. In fact, most engineering objects are multi-input-multi-output systems. Although many attempts have been made, there is no way to design such systems with classical control theory.

To a satisfactory result;


3. Classical control theory adopts heuristic method to design the system, mostly relying on feedback mode for control.



Modern control theory: Modern control theory uses linear algebra and differential equations as the main mathematical tools, based on the state space method, to analyze and design control systems.


Modern control theory has the following characteristics:


1. The transformation of the structure of the control object The structure of the control object is changed from a simple single-loop mode to a multi-loop mode, that is, from single-input single-output to multiple-input multiple-output. It must deal with the optimization and control of extremely complex industrial production processes.


2. Transformation of research tools

1) Integral transformation method is transformed to matrix theory and geometric method, and from frequency method to state space research;

2) The development of computer technology has shifted from manual calculation to computer calculation.


3. The transformation of modeling methods has changed from mechanism modeling to statistical modeling, and statistical modeling methods of parameter estimation and system identification have been adopted.


The further development of modern control theory includes the following aspects:

1) Modeling and system identification

2) Optimal control theory

3) Adaptive control theory


Intelligent control theory: has the following distinctive features:


First, when analyzing and designing intelligent control systems, the focus should not be placed on the analysis and design of traditional controllers, but on the smart machine model, which means that the focus should not be placed on mathematical formulas.

The description, calculation and processing! In fact, some complex large systems may not be described by precise mathematical models at all, but the focus should be on the description, symbols and environment of non-mathematical models.

Identification, knowledge base and inference engine design and development, etc. come up.


Second, the core of intelligent control is high-level control, and its task is to organize the actual environment or process, that is, decision-making and planning, to achieve generalized problem solving.


Thirdly, intelligent control is a borderline interdisciplinary subject. Professor Fu Jingsun first proposed the binary intersection theory of intelligent control, that is, the intersection of artificial intelligence and automatic control. In the United States, Celides and 1977

Extend Fu Jingsun’s binary structure to a ternary structure! That is, the intersection of artificial intelligence, automatic control, and operations research. Later, Professor Cai Zixing of Central South University of Technology expanded the ternary structure to a quaternary structure!

The intersection of artificial intelligence, automatic control, operations research and information theory further improves the structural theory of intelligent control.


Fourth, intelligent control is an emerging research and application field with extremely attractive development prospects. Since the concept of “intelligent control” was put forward to the present, automatic control and artificial intelligence experts and

Scholars have proposed various intelligent control theories, and some of them have played an important role in practice.

In the long talk above, it is completely based on the division of the theoretical field. In practical engineering, often you have me in you and you in me. It is difficult to simply define. The core is that control needs to be transferred from the traditional simple

Only when the recursive function is developed to model with big data can it be controlled more accurately.


The device bank is the pioneer in this area in China. You can search for "device bank" in the WeChat applet to experience the minimalist cloud access process. All software is free.

2. Current status

At present, in domestic industrial control, whether it is metallurgy, electric power, equipment, building, most control systems are implemented, we need to decompose each complex control system into one

This method has been used in the industry for decades. It is simple, reliable, and effective. A basic PID plus some fuzzy control can be used.

To meet most of the requirements.

There are a series of problems in the realization of this kind of automatic control engineering:

1. The parameters are fixed: the control parameters are determined during the system debugging, mostly based on experiments and experience. In the process of system operation, it is no longer modified. In actual operation, this

These control parameters are not necessarily suitable, which leads to the inefficient operation of the system. 2. Poor adaptation: The economic range of the control function is narrow, because all control algorithms have a range. A good control algorithm can meet the performance requirements of the user in the full range, but in most cases, it may not be able to It is economical. Of course, in practice, there are also some capable engineers segmenting the algorithm, and different control intervals use different algorithms or parameters. Generally speaking, the same control algorithm, debugging by different people, may cause the high-efficiency range of the system to be in different positions. In fact, we often see a large number of systems operating in the low-efficiency range for a long time in the field. 3. Too many human factors: Whether it is the choice of algorithm, the fine control, and the choice of parameters, they are all closely related to the engineer's ability and luck and are not controlled. For example, the simplest heating may have a small amount of discharge during commissioning. The project can meet the requirements during commissioning and acceptance evaluation, but when it is officially put into operation, it will be controlled because of the increase in load. The effect becomes worse and the fluctuation increases. 4. Unable to evaluate effectively: After the system is put into operation, most of the systems will not be evaluated again. For example, the energy consumption in building control, we found that the energy consumption of many old systems is several times the theoretical value. But because there is no effective evaluation, millions of electricity bills are wasted year after year.

Three, a new generation of control

In traditional control, the algorithm is determined and the parameters are fixed during the system construction period. Under this circumstance, a large number of automatic control systems are running in the low-efficiency range, and because there is no evaluation system, no one knows, and finally a large amount of energy consumption is wasted, and the quality of the products produced is unstable. If the traditional control algorithm is used again, these problems are difficult to solve. We can only add more judgments and more segments manually, and use different algorithms or parameters for different intervals.

Modern control theory and intelligent control are essentially to solve these problems and solve these non-linear, multi-input and multi-output systems. But the core of modern control theory is state space

Intermediate method is modeling, which requires a lot of data to describe. Just as the most popular artificial intelligence deep learning algorithm, its core is massive data for training to make it self

Find the feature and then identify it.

Compared with the IT industry, automatic control needs to be more precise and simple, because the information for automatic control is limited, and the selection of its characteristics (input information) is also artificial, which requires us to install sensors in specific locations to get it. Therefore, the big data in the automation industry is pre-modeling and precise characteristics. What is needed is to classify and summarize the data, and get its correlation.

At present, artificial intelligence in the IT industry focuses more on the analysis and reasoning of people, vision, language, and network data, which is technically more complicated.

For the automation industry, this technology is not appropriate. It is difficult for us to randomly install a large number of sensors in the traditional system, and then analyze the internal correlation from the data. In fact, it is completely unnecessary. In the automatic control industry, we must artificially estimate in advance which parameters are related to the control effect, and then collect this information and send it to the control system for analysis.

In traditional island-style control systems, you can only use your own historical data for analysis. These data are misleading and may not even be convergent, even if they are effective.

The process is too long, which is the core reason why the early expert database system could not be promoted.

To give a simple example, a central air-conditioning system in a building control has two core indicators, one is comfort and the other is energy consumption. In fact, we can compare different energy consumption of wind or water, and

To achieve the same comfort effect, either the air volume is small, the water volume is large, and the temperature difference between the inlet and return water is small, or the air volume is large, the water volume is small, and the temperature difference between the inlet and return water is large. The energy consumption of the two will be quite different, and the equipment life will also have a significant difference.

In the current control system, the entire central air conditioner is generally decomposed into a wind system, a chilled water system, a chiller system, and a cooling water system. Each system is a single input or dual input

In a single-output system, there is no correlation between the subsystems, which is a typical loose coupling. The advantage is that each subsystem has its own indicators, debugging is simple and convenient, and the system stability is high, which is an asynchronous system in an electronic system. However, the disadvantages of this method are quite obvious, that is, it is difficult to achieve the optimum and cannot meet the performance requirements of modern control.

As an on-site engineer, he does not evaluate energy consumption indicators and system life. He only needs to satisfy the user’s comfort level. So the final result is that different people use different procedures to achieve the same

This kind of comfort has a difference of tens of percent in energy consumption, and the life span of the main engine and water valve will also be greatly affected.

Another example: a building control system, no one checked in during the commissioning, and after the commissioning was completed, the building was built into a supermarket, crowded with people every day, the initial parameters were not suitable, causing comfort

With a moderate decrease, energy consumption has also greatly increased, but the ability of users to maintain electricians cannot be understood and changed.


So how to solve these problems?
This brings us back to modern control theory, we need to use enough data, a reasonable modeling, and iterate out the relationship between variables. The specific implementation is as follows:


The first method is for specific control systems, especially process control systems. After the project is implemented, control-related variables are collected and uploaded to the server through the network. The server summarizes, analyzes and evaluates all the same types of control objects in the world, and obtains For the optimal control parameters under current conditions, directly download a set of optimal parameters derived from cloud computing.


The difference between the second method and the first method is that the cloud directly participates in the control, as the outermost large closed loop with large delay, for trend control.


The significance of this system:

1. The control system is no longer the completion of the project, the control system company handed over the keys and it was finished. But from the end of the project, its control parameters will continue to be analyzed based on big data during operation.

Analysis and calculation make it more and more optimized and have the strongest adaptability.


2. The nature of automation companies has changed. Automation companies no longer focus on engineering, but on maintenance and management services. Each industry only needs a few automation companies, and they will be this

The most comprehensive service provider with the largest amount of data in the industry. After the automation system is connected to the company's cloud platform, it will dynamically calculate and optimize. On the one hand, the automation company charges engineering fees and further charges

Taking the optimization service fee, the more important thing is that users no longer need to manage the equipment in the future, but are managed by the automation company, which derives automated insurance or leasing business.


3. The production efficiency and benefits of the entire industry have been greatly improved. Traditionally, an automation owner needed to raise a group of second-rate maintenance engineers. Now it is no longer needed, but the automation company

During maintenance, a large number of maintenance personnel will be unemployed. On the other hand, the energy consumption of operation is greatly reduced and the quality is greatly improved. This is especially obvious in many industries with relatively high energy consumption.


4. With its own evaluation system, users will clearly know the efficiency and quality of their system in the same type of system, so that they can make more accurate system upgrade investments.

Four, adapt to


In the long run, with this control model, when the communication cost is approximately zero in the future, all control systems will have network access points, which may lead to major changes in the entire control industry.


But our automation industry does not need extremists. The introduction of any new technology requires a long development process. For now, the ones that can be used immediately must have the following characteristics:


1. Large amount: there are enough samples.


2. Valuable: Either it can save people, or it can save energy, or it can improve product quality.


3. Short payback period: After investment, you can pay back the cost within one or two years.


4. Security: This system is network-based. Although it is possible to divide the network of people and equipment through the encrypted communication process, users confirm and download control parameters and other means, but after all, this is an Internet-based big data system, and security is the first of. Therefore, we have to choose to bring our own security, even if there is a communication failure, there should be no security impact due to malicious hacker attacks.


5. Concluding remarks The Internet of Things, expert database systems, neuron networks, deep learning, and artificial intelligence, these early thoughts in the theoretical and engineering circles will finally come over. It will be due to big data and cloud computing. Get a new life, and finally vigorously change the entire control field, and then deeply affect the entire industry and human life. In the future, a large number of industries will be changed as a result, and a large number of employees will be diverted. This is by no means unfounded, but an immediate need to face.


In this new industrial revolution, most people focus on the black lamp factory and focus on Industry 4.0. In fact, these are just appearances, and the core is that control theory will produce qualitative changes. Future control will be like snowballing from the top of a mountain. Firstly, it is controlled by classical theory and the initial data is obtained. Then other control objects no longer need control algorithms, but are based on a data matrix corresponding to big data. No matter what kind of control, in theory, it can be based on a sufficiently large search The table method is used for control, and the cloud will also evaluate and adjust according to the operation situation.


It is interesting to think about it. In the future automation system, you only need to tell it that this is a blast furnace, what capacity, what type of mine is used, and where it will generate a set of data. The estimated control system calls the lookup table output. The only thing the control system needs to do is to smooth the steps between the table data.


Looking further away, the looming behemoth makes people a little yearning and awe-inspiring. There will be a lot of lives in the entire industry that will change. As a leading domestic automation company, Rectangular has the responsibility to develop seriously and responsibly and apply these technologies carefully.


In the past ten years, I have been producing products related to the Internet of Things, and using these control systems to help users complete various projects, from early street lamp monitoring, sewage monitoring, energy management, smart buildings, digital oil fields, digital agriculture to The current CDB system is a process of accumulating and thinning. The new generation of control system architecture, such as the throat, has to be sent and can no longer be suppressed. From a selfish point of view, it would be more beneficial to close the door and slowly develop, but I am a lazy person and need more colleagues to enter to stimulate and help improve the industry. Therefore, I have published this article without being exhaustive. I believe that there will be a lot of friends and friends who can understand it, and I hope that you can send and call more to communicate.


After the dry goods are blown, I will take some private goods.


Device Bank: It is a set of development tools for device cloud, which can realize device cloud in 30 minutes, no need to know programming. Use the mobile phone to search the small program "device bank", which can be used for free without registration. This is also the reason why I have to abandon the IC design career of tall and large and invest in the red sea of ​​industrial control. Everyone puts forward more opinions and forwards more. This is also the best way to help China's automation industry.


The article is original, please indicate the source for reprinting!

Public number: Laogou Technology (LG123321yun)






UV Vacuum Curing Machine

Uv Vacuum Curing Machine,Uv Protection Screen Protector,Uv Glass Curing Screen Protector,Uv Film Screen Protector Phone

Shenzhen TUOLI Electronic Technology Co., Ltd. , https://www.szhydrogelprotector.com