AUTOMATION TECHNOLOGY:TECHNOLOGIES OF ARTIFICIAL INTELLIGENCE

By -

TECHNOLOGIES OF ARTIFICIAL INTELLIGENCE

Automation technologies have been bestowed intelligence by the invention of computers and the evolution of artificial intelligence theories. Because of the introduction of the technologies of artificial intelligence, automated systems, from the perspective of control, can intelligently plan, actuate, and control their operations in a reasonable time limit by handling / sensing much more environmental input information (see the horizontal axis in Figure 3). Meanwhile, artificial intelligence increases the decision making complexity of automation technology, but the cost of the system that is automated by the technologies of artificial intelligence is relatively low compared with the system automated by the automatic control theory (see the vertical axis in Figure 3). Figure 3 shows a roadmap of how various automation technologies influence the development of automation systems in two axes. In the following subsections, those technologies of artificial intelligence, including neural networks (NN), genetic algorithms (GA), knowledge-based systems (KBS), fuzzy control, and the hybrid of the above-mentioned technologies will be introduced in general.

Knowledge-Based Systems

The structure of a knowledge-based system (KBS) in control is presented in Figure 4. The control decisions are achieved by reasoning techniques (inference engine) rather than quantitative computa- tion and can deal with uncertainties and unstructured situations. The knowledge base is updated by continuously acquiring the knowledge from experts, the decisions made by the operator and the feedback from the process. Nowadays, KBSs have been applied in many fields to diagnose causes of problems, such as in medical diagnosis, vehicle troubleshooting, and loan strategy planning in banks.

Automation Technology-0007

Automation Technology-0008

The numerous industrial engineering applications of control models in computer information systems can be classified into two types: (1) development of decision support systems, information systems that provide the information, and decisions to control operations, and (2) maintenance of internal control over the quality and security of the information itself. Because information systems are usually complex, graphic models are typically used.

Any of the control models can essentially incorporate an information system, as indicated in some of the examples given. The purpose of an information system is to provide useful, high-quality information; therefore, it can be used for sound planning of operations and preparation of realistic standards of performance. Gathering, classifying, sorting, and analyzing large amounts of data can provide timely and accurate measurement of actual performance. This can be compared to reference information and standards that are also stored in the information system in order to immediately establish discrepancies and initiate corrective actions. Thus, an information system can improve the control operation in all its major functions by measuring and collecting actual performance measures, analyzing and comparing the actual to the desired set points, and directing or actuating corrective adjustments.

An increasing number of knowledge-based decision support and control systems have been applied since the mid-1980s. Typical control functions that have been implemented are:

• Scheduling

• Diagnosis

• Alarm interpretation

• Process control

• Planning

• Monitoring

Artificial Neural Networks

The powerful reasoning and inference capabilities of artificial neural networks (ANN) in control are demonstrated in the areas of

• Adaptive control and learning

• Pattern recognition / classification

• Prediction

To apply ANN in control, the user should first answer the following questions:

1. If there are training data, ANN paradigm with supervised learning may be applied; otherwise, ANN paradigm with unsupervised learning is applied.

2. Select a suitable paradigm, number of network layers, and number of neurons in each layer.

3. Determine the initial weights and parameters for the ANN paradigm.

A widely applied inference paradigm, back propagation (BP), is useful with ANN in control. There are two stages in applying BP to control: the training stage and the control stage (Figure 5).

Training Stage

1. Prepare a set of training data. The training data consist of many pairs of data in the format of input–output.

2. Determine the number of layers, number of neurons in each layer, the initial weights between the neurons, and parameters.

3. Input the training data to the untrained ANN.

4. After the training, the trained ANN provides the associative memory for linking inputs and outputs.

Control Stage

Input new data to the trained ANN to obtain the control decision. For instance, given a currently observed set of physical parameter values, such as noise level and vibration measured on a machine tool, automatically adapt to a new calculated motor speed. The recommended adjustment is based on the previous training of the ANN-based control. When the controller continues to update its training ANN over time, we have what is called learning control.

Other application examples of neural networks in automated systems are as follows:

• Object recognition based on robot vision

• Manufacturing scheduling

• Chemical process control

The ANN could be trained while it is transforming the inputs on line to adapt itself to the environment. Detailed knowledge regarding the architectures of ANN, initial weights, and parameters can be found in Dayhoff (1990), Freeman and Skapura (1991), Fuller (1995), Lin and Lee (1996).

Fuzzy Logic

Figure 6 shows a structure of applying fuzzy logic in control. First, two types of inputs must be obtained: numerical inputs and human knowledge or rule extraction from data (i.e., fuzzy rules). Then the numerical inputs must be fuzzified into fuzzy numbers. The fuzzy rules consist of the fuzzy membership functions (knowledge model) or so-called fuzzy associative memories (FAMs). Then the

Automation Technology-0009

Automation Technology-0010

FAMs map fuzzy sets (inputs) to fuzzy sets (outputs). The output fuzzy sets should be defuzzified into numerical values to control the plant. Due to the powerful ability of fuzzy sets in describing system linguistic and qualitative behavior and imprecise and / or uncertain information, many indus- trial process behavior and control laws can be modeled by fuzzy logic-based approaches. Fuzzy logic has been applied in a wide range of automated systems, including:

• Chemical process control

• Autofocusing mechanism on camera and camcorder lens

• Temperature and humidity control for buildings, processes, and machines

Genetic Algorithms

Genetic algorithms (GAs), also referred to as evolutionary computation, are highly suitable for certain types of problems in the areas of optimization, product design, and monitoring of industrial systems. A GA is an automatically improving (evolution) algorithm. First the user must encode solutions of a problem into the form of chromosomes and an evaluation function that would return a measurement of the cost value of any chromosome in the context of the problem. A GA consists of the following steps:

1. Establish a base population of chromosomes.

2. Determine the fitness value of each chromosome.

3. Create new chromosomes by mating current chromosomes; apply mutation and recombination as the parent chromosomes mate.

4. Delete undesirable members of the population.

5. Insert the new chromosomes into the population to form a new population pool.

GA are useful for solving large-scale planning and control problems. Several cases indicate that GA can effectively find an acceptable solution for complex product design, production scheduling, and plant layout planning.

Hybrid Intelligent Control Models

Intelligent control may be designed in a format combining the techniques introduced above. For example, fuzzy neural networks use computed learning and the adaptive capability of neural networks to improve the computed learning’s associative memory. Genetic algorithms can also be applied to find the optimal structure and parameters for neural networks and the membership functions for fuzzy logic systems. In addition, some techniques may be applicable in more than one area. For example, the techniques of knowledge acquisition in KBSs and fuzzy logic systems are similar.

Comments

Post a Comment

Popular posts from this blog

MATERIAL-HANDLING SYSTEMS:STORAGE SYSTEMS

By -
Image
STORAGE SYSTEMS Various types of racks have been developed to store loads of different sizes and shapes. One of the simplest storage systems is block stacking, where loads (which may or may not be palletized) are simply stacked on top each other with no separate support structure. While block stacking is inex pensive and provides dense storage (i.e., little or no space is lost due to air between the loads), its use is limited in practice because only certain types of loads (such as appliances or furniture packed properly in cartons) can be stacked without crushing / damaging the loads at the bottom and access to individual loads (i.e., load selectivity) is very limited. In fact, loads must be removed from a stack, starting with the load at the top. This leads to last-in-first-out (LIFO) stock rotation, which is not desirable for many storage systems. A fundamental trade-off that exists in storage systems design is that between storage density (how well the cubic volume is utilized) ...
Read more

NETWORK OPTIMIZATION MODELS:THE MINIMUM SPANNING TREE PROBLEM

By -
Image
THE MINIMUM SPANNING TREE PROBLEM The minimum spanning tree problem bears some similarities to the main version of the shortest-path problem presented in the preceding section. In both cases, an undi r ecte d and connected network is being considered, where the given information includes some mea- sure of the positive length (distance, cost, time, etc.) associated with each link. Both problems also involve choosing a set of links that have the shortest total length among all sets of links that satisfy a certain property. For the shortest-path problem, this property is that the chosen links must provide a path between the origin and the destination. For the minimum spanning tree problem, the required property is that the chosen links must provide a path between ea c h pair of nodes. The minimum spanning tree problem can be summarized as follows: 1. You are given the nodes of a network but not the links. Instead, you are given the po- tential links and the positive length for each ...
Read more

DUALITY THEORY:THE ESSENCE OF DUALITY THEORY

By -
Image
THE ESSENCE OF DUALITY THEORY Given our standard form for the primal problem at the left (perhaps after conversion from another form), its dual problem has the form shown to the right. Thus, with the primal problem in maximizatio n form, the dual problem is in minimization form instead. Furthermore, the dual problem uses exactly the same pa r amete r s as the primal problem, but in different locations, as summarized below. 1. The coefficients in the objective function of the primal problem are the right-hand sides of the functional constraints in the dual problem. 2. The right-hand sides of the functional constraints in the primal problem are the coef- ficients in the objective function of the dual problem. 3. The coefficients of a variable in the functional constraints of the primal problem are the coefficients in a functional constraint of the dual problem. To highlight the comparison, now look at these same two problems in matrix notation (as introduced at the beginn...
Read more