Karen Kosky ABE 565 Expert System Article Review Week 7 10-4-95Ross, John; "An Expert System for Soil Erosion Mitigation in Logging Operations on Steep Land;" AI Applications; Vol. 7, No. 4, 1993; pp.69-70.
The author describes a problem of soil erosion which occurs on the logged land of New South Wales, Australia. Erosion is of particular concern for local residents due to logging on steep slopes and environmentally sensitive areas. The Department of Conservation and Land Management (CaLM) in New South Wales developed a comprehensive set of restrictions for future logging operations. Before logging, the logger must apply for authority to remove or destroy trees. An officer of CaLM must inspect the land to ensure that logging would not have a significant detrimental effect on the environment.
As an addition to the development of the restrictions and rules, CaLM created an Expert System to simplify the process of assessing the land parcels. The program was developed in the Level5 Object shell. The knowledge base is split into two areas; one for common environmental conditions (slope, soil erodibility, climactic conditions, etc.) and one for other special conditions (duration of logging operation, special landscape features, etc.) The rule base was developed by interviewing three field experts who regularly carried out inspections of protected land.
The benefits which could be realized from such a system were listed as: * consistency in decision making in Forestry operations statewide * access to expertise for less experienced field workers * development of a standard for erosion mitigation * decision-making capability on-site, and * easy modification capabilities for adaptation to other tree removal activities.The expert system was slated to be in use in the field at the end of 1993.
Summary of Journal Article Bruce Worstell Expert System for Selecting Conservation Planting Machines: 'PLANTING' John E. Morrison and others Transactions of the ASAE, v. 32(2): March-April 89 p. 397Conservation tillage requires a specific combination of implements for a given no-till crop system. Implement companies have produced no-till planters and drills with a wide variety of options available to growers. Growers must consider many aspects of their resources in order to correctly identify the appropriate conservation planter or drill. To aid growers, an expert system was developed to help identify the appropriate planter based on farm location, soil type, slope, crop rotation, and type of tillage.
The "PLANTING" expert system was developed using software from EXSYS Inc. Users are prompted to input data related to soil conditions and cropping practices and run the model. The model analyzes the input selections to make sure they do not exceed the allowable soil loss. The ES then reviews the rules and recommends the appropriate components for a planter and models that currently exist with that configuration.
Facts are necessary in order to develop rules that can be evaluated by the inference engine. The soil data for this study was obtained from the Soils-5 Interpretation and other related databases. This data was then incorporated into a site specific database so that additional information such as rainfall could be recorded.
Allowable soil loss is determined using the Universal Soil Loss Equation. Facts about the soil, crop, residue, and operational parameters are incorporated into rules that determine if the cropping system keeps erosion below the tolerable level. After the soil loss restrictions are met, the ES identifies a planting machine.
Planters are described in the ES using attributes such as residue cutting, soil cutting, and furrow closing. Rules were developed based on facts about each component and how it engages the soil. The ES then selects an appropriate configuration and identifies existing machines on the market.
The grower receives a final report that describes the components and existing implements that would be most appropriate for the specified cropping system. The grower can then choose to buy new planter or alter the components on an existing planter according to the report.
Review : Advisory Expert System for Flexible Pavement Design Submitted by Chung-I Kevin Yen Source : Artificial Intelligence in Engineering n.8 1993, p.47-56 Author : A.T.C Goh
This paper describes the prototype of expert system PAVEDKB ( Pavement Design Knowledge Base) for flexible pavement design. The tasks for flexible road pavements design are often very complex due to the large number types of possible algorighm, pavement materials and traffic data. There are several methods that can be used in designing flexible road pavements. Some of the methods are empirical or experience based. The domain knowledge of PABEDKB are based on the design method developed from NAASRA (National Association of Australian State Road Authorities).
In the first part of this paper, the design method of NAASRA is introduced. The architecture of PAVEDKB is described in the second portion. PAVEDKB is developed on a backward chaining expert system shell, CRYSTAL which runs on IBM compatible microcomputers. PAVEDKB is designed as an highly interactive system in which the users are required to answer a series of question. PAVEDKB consists of four submodules, namely, the determination of the subgrade, the pavement material, the traffic loading and the pavement analysis. Each module was developed separately and can be accessed from the main menu. When executing the first three ( subgrade, pavement material and traffic loading) modules, PAVEDKB will interact with users to collect the site specific data and finally suggest a design result. If the site specific data are unabailable, PAVEDKB will provide subjective guidelines for the characterization of the subgrade, pavement materials and traffic data. When running the pavement analysis which is commonly determined by using the FORTRAN computer program CIRCLY, the knowledge-base will activate the CIRCLY program and extract the output result. In addition, the explanation facility embedded inCRYSTAL can respond to queries as to why certain results are derived as well as to provide a trace of the inference login. The inference net of all modules are shown in this portion.
The third portion of this paper are some sample rules and dialog screens and running results of each module.
The best benifit of PAVEDKB is to help those local highway agencies lacking human experts. The NAASRA design methodology for pavement design relies on the experiences of CIRCLY program which grnerally requires special expertise and are not user friendly. PAVEDKB currently is limited to the designof flexible pavements. It will be extent to refining the knowledge base to incorporate rigid pavement design and optimal design algorithm in the future.
Soybean Oil Extraction Diagnostic Expert System Applied Engineering in Agriculture Vol8(4):Jul 92
Soybean oil extraction is a very difficult process. There is an optimization of small flakes for extraction oil efficiency and the fact that they clog the system. The final product SOYEX was developed to aid in the processing of soybean oil. The expert system uses a series of rules of if then statements to give conclusions. The Knowledge Engineering System (KES) was used to develope the system. The system was then converted for use on several computer system.
Week 6 Assignment- Article on Expert Systems Title: Knowledge integrating system for the selection of solvents for extractive and azeotropic distillation. Source: Computers and Chemical Engineering, Vol 18 Supplement p s25-29 (1994)
Often a third solvent is used in extractive and azeotropic distillation to give a greater difference in the relative volatilities between the two solvents therefore allowing easier separation of components. The selection of these seperating agents requires research or expert knowledge to select something thatwill give a good performance and also be easy to seperate at a later stage.
This article implemented SOLPERT - solvent selecting expert system - which is able to use databases, heuristics and empirical forms of knowledge and rules to perform the required selection. Solpert works in 4 stages; selection of classes selection of chemically similar groups, propose solvents and then rank the proposed group of solvents. The first stage classifies the feed solvents using Berg et al classifications which primarily uses chemical types and functional groups. Selection of similar chemical groups is done by utilising various databases depending on the characteristics of the feed solvents. The third and fourth steps are able to sort out possible agents by considering possible azeotropes and ease of further separation at a later stage. The way the steps are sorted out a large number of options and alternatives are eliminated by the first and second steps leaving the third and fourth steps to perform the finer work.
This tool will allow the selection of solvents to be performed at a faster rate eliminating a lot of book work that would otherwise be required.
Artificial Intelligence and Expert Systems in Agricultural Research and Education. Barrett, J.R., J.B. Morrison, and L.F. Huggins. 1985 Winter Meeting American Society of Agricultural Engineers, paper no. 85-5516.
This article gave a good overview of the history of artificial intelligence and expert systems as well as a look into the future applications of AI. The most interesting part dealt with the agricultural applications. One aspect that was particularly attractive of AI to the agricultural experts was Natural Language understanding of computers together with voice recognition. Natural language capabilities was of particular interest when it was linked to the following three applications:
1. Voice control of agricultral machinery 2. Increase ease of computer use in management of farm business 3. Increase in ease of record keeping and data collection.Robotics was also of great interest to agriculture experts. They believe that robots can be important in agriculture to replace humans in labor intense tasks, to improve quality of tasks, to improve safety, and to improve work flow (self-diagnosing machines to reduce repair time during peak times such as planting or cultivation). Finaly it was felt that expert systems should be applied to agriculture in at least two major areas, decision support and troubleshooting. It was suggested that the priority to developing expert systems should be in marketing support systems, pest management systems, and troubleshooting and diagnosis expert systems for agricultural machinery. In the agricultural areas of science and education it was believed that ES could be applied to resource management, financial management, program evaluation, and personnel management. Research situations would find ES useful for control of laboratory equipment, organization of data, classification, and statistical design and experiment. Even though this article was written 10 years ago, it still portrays the importance of AI in agriculture and the need to apply natural language understanding, robotics, and expert systems to the benefit of agriculture.