China
Africa
Explore chapters and articles related to this topic
Integration of Forward and Backward Inferences Using Extended Rete Networks
Published in Takushi Tanaka, Setsuo Ohsuga, Moonis Ali, Industrial and Engineering Applications of Artificial Intelligence and Expert Systems, 2022
Most current knowledge engineering tools have been designed to provide the structures for representing domain knowledge by using rule structure (IF-THEN), and an inference mechanism for controlling the use of domain knowledge. In production rule systems, two very important reasoning control strategies are forward (data-directed) chaining and backward (goal-directed) chaining [HWB83]. Although forward and backward chainings are useful inference schemes, forward chaining approaches sometimes have the disadvantage of generating many facts not directly related to the problem under consideration, while backward chaining approaches have the disadvantage of perhaps becoming fixed on an initial set of goals and having difficulty shifting focus when the data available do not support them. An interesting possibility is to inference in both directions simultaneously or opportunistically. We call this inference mechanism as mixed inference. To remedy control limitations within the basic architecture, several works have been tried. Some knowledge engineering tools (ART, TIRS, ESE, KEE, etc.) can process both forward and backward chaining [JC94]. However, their chaining schemes are not be naturally integrated tightly into a single uniform control framework.
The Electronic Encapsulation of Knowledge for Groundwater Quality Management
Published in Nebojša Kukurić, Development of a Decision Support System for Groundwater Pollution Assessment, 2020
All the methods and techniques for knowledge representation developed within AI, with the exception of a case-based reasoner, are rule-based. Even the most recent approach that introduces a concept of intelligent agents (Russell and Norvig, 1995), does not depart from the traditional rule formalism. The rules are derived using the principles of logic, i.e. syntax, semantics and deduction14. So-called production rules (simply ‘If-Then statements) are the most widely-used structures for knowledge representation. The inferences are made by using forward or backward chaining. Rules are applied in forward chaining to derive new facts from those that are already known. In backward chaining, a conclusion is made in the first place; then recursive inferencing is applied to define whether conditions are met that allow such conclusion. If set of rules condition a situation (or an action to be taken), it is convenient to combine those rules in so-called decision tables (Table 2.2). The rules and their organisation are comprehensively described in a number of books (e.g. David and King, 1977, Weiss and Kulikowski, 1984).
Evaluation and Use of Clips for Developing Temporal Expert Systems
Published in Don Potter, Manton Matthews, Moonis Ali, Industrial and Engineering Applications of Artificial Intelligence and Expert Systems, 2020
Susan J. Chinn, Gregory R. Madey
Our requirements for inference engines that can support temporal representation and reasoning include the ability to use both forward and backward inferencing strategies and the ability to support truth maintenance. Forward chaining is well suited to data-driven applications, such as monitoring which uses temporal data. Backward chaining algorithms, often used in planning and diagnostic applications, use time as part of the consultative process to determine when events occurred, or the duration of certain states. Shells that combine both methods of inference could be extremely useful in planning problems, such as providing a temporal sequence in which goals could be solved [PA90]. Truth maintenance in temporal applications involves the “frame problem,” where the system may need to determine which facts have not changed as events occur over time [MH69]. The assumption that facts will remain “true” unless explicitly changed, known as persistence, allows previous assertions to continue to be true unless contradictory information is received later in the reasoning process [McD82]. Truth maintenance also requires support for non-monotonic reasoning, so that facts that become false can be retracted.
Function block-enabled operation planning and machine control in Cloud-DPP
Published in International Journal of Production Research, 2023
Mohammad Givehchi, Yongkui Liu, Xi Vincent Wang, Lihui Wang
In the current research, the core of the operation planner unit is designed as a production rule system in which rules are the main knowledge building blocks. The rule-based inference engine goes through different stages of the decision-making flow and uses a forward-chaining technique to activate necessary rules and infer desired operation plans. Figure 2 shows two different rules flows with different numbers of stages. The first one is for the cases where rules are mainly defined in context of an explicit strategy and hence different operations related to different steps are created and completed through the next stages. This flow is suitable for the cases where a set of alternative parametric operation plans or strategies are suggested for a conditioned feature. The second flow is based on the rules representing granulated and atomic single step strategies that convert one input MF to an output MF. The volumetric feature and a selected aperture surface set of the input MF must be encompassed in the ones of the output MF with the same selected TAD (Refer to (Givehchi, Haghighi, and Wang 2015b) for the definitions). At runtime, different suitable steps would get chained together dynamically to form an operation plan. While the first approach might be more comprehensive for operation planners and knowledge managers, it suffers from knowledge redundancy, which will result in difficulties and errors in change management and little reuse of existing knowledge.
Prognostic modelling for industrial asset health management
Published in Safety and Reliability, 2022
Neda Gorjian Jolfaei, Raufdeen Rameezdeen, Nima Gorjian, Bo Jin, Christopher W. K. Chow
Expert Systems are the most popular knowledge-based techniques that require less training time than machine learning techniques. An ES is an intelligent computer program that simulates the knowledge and inference procedures of human experts to solve complex problems (Bollinger & Smith, 2001). It works well when the problem to be solved is complicated and when judgement and experience are helpful tools in finding the solution. An ES has two essential parts: the knowledge base and inference engine. Acquired knowledge is put into the rule base or knowledge base in an ES. Rules and facts make up the knowledge base. Rules in ESs are defined using IF-THEN syntax, where IF is an antecedent that is compared to inputs and THEN is a consequent, which is the output (Sharma et al., 2008). The inference is the derivation of new facts from known facts. There are two inference approaches that are commonly used as problem-solving strategies in ESs, forward and backward chainings. Forward chaining is reasoning from facts to conclusions resulting from those facts. Backward chaining involves reasoning in reverse from a hypothesis to facts that support a hypothesis. An inference engine is the processor of facts. The inference engine within an ES is the portion of the computer program that processes knowledge.
A knowledge-based document assembly method to support semantic interoperability of enterprise information systems
Published in Enterprise Information Systems, 2022
Marko Marković, Stevan Gostojić
The inference engine tests if a conclusion is true or false by using a forward-chaining or backward-chaining algorithm (Sharma, Tiwari, and Kelkar 2012). The proposed solution in this work leverages a DR-DEVICE reasoner based on a forward-chaining algorithm (Kontopoulos, Bassiliades, and Antoniou 2011), which infers all conclusions from the rules in the rule base and exports results. The conclusions are stored in RDF format using a simple ontology defined for the DR-DEVICE reasoner. The proof is stored in an extended RuleML format that supports DR-DEVICE proof capabilities.