Expert system is a type of computer applicati on program that makes decisions or solves problems in particular field, by using knowledge and a nalytical rules defined by expert in the field. UTILIZATION OF KNOWLEDGE-BASED EXPERT SYSTEMS. TRAFFIC MONITORING PROGRAMS – A FOCUS ON TRAFFIC PATTERN GROUP ANALYSIS by Abdulkadir Ozden A dissertation submitted to the Faculty of the University of Delaware in partial. 5.2 An Expert System Development Tool: Exsys Corvid Core.
An expert system for automating sandstone acidizing has been developed in this study. The system consists of six stages, which were built following an acidizing logic structure that is presented in the form of decision trees. The six stages consist of formation oil displacement, formation water displacement, acetic acid, HCl pre-flush, main acid, and over-flush stage. The acid blends recommended by the system are damage-type specific, and account for the compatibility between the injected acid and the in situ crude in order to avoid formation of asphaltene sludge, or emulsions. The acidizing expert system has been implemented as an online web-based application. Applicability of this expert system to acidizing design has been illustrated using three documented actual field cases spanning the Niger Delta region, Algyo Oil field in Hungary, and the Dulang oil field in Malaysia.
For Niger Delta field and the Algyo field cases the expert system produced an optimal main acid job design with recommended pre- and post-flushes that are in perfect agreement with successful field treatment. For the Dulang oil field, in actual practice, an organic clay acid was injected for removing problems of fines migration in a reservoir that has a high calcite content, with a moderate amount of feldspar and chlorite clay. The acidizing expert system recommended a chelant-based acid, which is a recent innovation that is considered a more cost-effective acid solution for dissolving fines in presence of calcite and other sensitive clay minerals. The selection of an appropriate acid type, concentration and volume needed to be injected along with the required additives and their concentrations for various temperature and mineralogical environments can be a very perplexing task.
Part of this problem stems from the complex and heterogeneous nature of most sandstone rocks. In addition, the interactions between the many different mineral species and the injected acid depend not only on their chemical compositions but also on temperature, and on surface morphology (Boyer ).Sandstone formations are challenging to acidize because of the presence of alumino-silicates such as clays, zeolites, and feldspars, which may lead to unwanted precipitates in contact with mud acids as a result of secondary and tertiary reaction products. For instance, smectite and mixed layer clays are unstable in HCl at temperatures of approximately 150 °F.
Chlorite is unstable in presence of HCl at temperatures above 125 °F. When contacted with HCl, the clay structure may disintegrate, releasing iron which may precipitate in presence of HCl acid (Rae and Di Lullo ). Therefore, formations with high levels of chlorite respond best to acid formulations based on acetic acid rather than hydrochloric acid, since the former limits iron liberation and thereby reduces the risk of precipitates from iron reaction products (Nasr-El-Din and Al-Humaidan; Hashem et al. In formations with high levels of feldspar (20%), a common practice is to limit the strength of HF acid stages to reduce the formation of complex fluorosilicate precipitates and other species that would result from excessive dissolution of the mineral by stronger acid (Coulter and Jennings ).Illite clays are troublesome when using HF acid due to the presence of potassium in this clay structure.
When dissolved, the potassium is readily available to react with the HF-alumino-silicate reaction products, forming the insoluble potassium hexa-fluosilicate. Illite is unstable in HCl at temperatures above approximately 150 °F. Kaolinite clay is considered the most detrimental from a migration standpoint (Coulter and Jennings ). It becomes unstable in HCl only at higher temperatures (greater than 200 °F).Zeolites are secondary minerals in the form of hydrated silicates of aluminum, calcium, sodium, and potassium. They are occasionally found in sedimentary rocks, with the most common form being analcite (analcime).
The knowledge and reasoning logic incorporated in the sandstone acidizing expert system take into account input data such as rock mineralogy, clay type and distribution, reservoir temperature, and formation fluids–acid compatibility. The treatment design is constructed following a sandstone acidizing structure that includes the following stages: Stage 1:Formation oil displacement Stage 2:Formation water displacement Stage 3:Acetic acid pre-flush Stage 4:HCl pre-flush Stage 5:Main acid injection Stage 6:Over-flushIn this structure, the pre-flush is no longer a single-stage, and may be stretched to multiple stages. In fact, individual Stages 1–4, or a combination of these may make-up the needed pre-flush stage, depending on the conditions of the rock after drilling, its mineralogy, presence of organic deposits, the formation water salinity, and the calcite content. The acidizing treatment design undertaken by Stages 1–6 is based on knowledge and experience extracted from human experts and arranged in hierarchical (tree) forms, which may be termed as decision trees which are used in the current work to represent the acquired knowledge and reasoning logic. A decision tree is a tree-like decision support tool that uses a graph to convey decisions and their possible consequences (Adamo; Yuan and Shaw ). It contains mainly three types of nodes: decision nodes associated with conditions and statements to support decision making, chance nodes to represent derived decisions and events that are likely to occur, and end nodes corresponding to situations and end goals to be obtained.
Decision trees are essential to understand and follow up the logic of the system. They may be considered as references and means of communication between the expert and the developer. Moreover, they facilitate maintaining, checking, modifying and extending the knowledge and logic of the system.Stage 1 addresses the cleaning of whole mud losses that take place during the drilling phase. It is also concerned with the cleaning of organic deposits in the formation. The decision tree for Stage 1 is given in Fig. This stage prepares the surfaces for the main treatment fluids.
Hydrocarbon solvents are used to clean oil films and paraffin deposits so that the main acid systems can contact the mineral surfaces. In Stage 2, brine containing ammonium chloride is used to help remove and dilute acid-incompatible species, such as potassium or calcium (Fig. ). This process helps avoid precipitation of some of the most detrimental precipitates produced in sandstone acidizing such as sodium and potassium fluorosilicates.
Ammonium chloride is also used to condition the clays that might come in contact with injected acids. The lower the formation water salinity is, the higher the concentration of ammonium chloride is needed to suppress the electrical double layer of clays.
A linear relationship between ammonium chloride concentration and water salinity has been adopted in this study. This is inspired from estimates of the critical salt concentration needed for clay stability (Schechter ). The boundary points of this linear relationship consist of 8% ammonium chloride solution for 0.1% water salinity, and 3% ammonium chloride solution for 5% water salinity, or greater.
Stage 3 is reserved for formations that bear iron-rich minerals, or iron-rich clays like chlorite (Fig. ). Injection of HCl acid in these rocks is likely to precipitate iron scales when iron-rich minerals are present (Coulter and Jennings ). In order to alleviate these problems, HCl is substituted with acetic acid when the volume fraction of these authigenic species is 6% (Fig. ). Acetic acid lessens the risk of precipitates from iron reaction products. Compatibility of the main acid with formation fluids is another consideration for pre-flushes.
A number of crudes may sludge in contact with certain acidic mixtures (Houchin et al. These situations may require buffering acetic acid as a pre-flush. In the absence of sludge and emulsion problems, HCl pre-flush (Stage 4) in sandstone acidizing becomes extremely important. The function of an HCl pre-flush is to remove the bacteria that may exist with injection wells, calcareous material growth in the pore system, or to remove CaCO 3 inorganic scale deposits, and the calcite cementing material that may precipitate calcium fluoride deposits in contact with HF acid of the main stage. HCl pre-flush also reduces the potential precipitation of insoluble or slightly soluble reaction products like calcium fluoride, and sodium and potassium hexafluorosilicates.
The decision tree for HCl pre-flush is shown in Fig. 1.Conventional mud acids are used only in very special circumstances and in general in low concentrations in order to avoid the precipitation of many damaging reactants, maintain formation integrity, and dissolve any fines. Indeed, mud acid is restricted for treating clay-clean formations at relatively low temperatures with insignificant amounts of calcite, Feldspar, or zeolite. 2.Organic acids are recommended for clean rocks that bear a minimum amount of clays, but at relatively high temperatures. The final design is reduced to include only necessary steps, though. A great emphasis is placed on Stages 4 and 5. In other words, Stage 1 is skipped if none of the diamond conditions of decision tree shown in Fig.
Are satisfied. Stage 3 is skipped, if the rock has no zeolite material, iron minerals, illite, chlorite, or mixed layer clays. Stage 4 is skipped if the conditions of the ‘OR-Test S4-1’, shown in Fig. a, are not satisfied. Decision trees that match the damage types and initial formation conditions for the six treatment stages have been constructed in this study. These decision trees make the logical foundation for the expert system decision-making process (Figs., ), covering the following damage types.
1.Particle damage from drilling and completion. 2.Fines migration.
3.Calcium carbonate scale. 4.Hydroxide scale (Mg(OH) 2, Ca(OH) 2).
5.Iron scales (FeS, Fe 2O 3, FeCO 3). 6.Polymer residue from drilling or secondary recovery. 7.Bacterial infestation (injection wells)In an effort to extend the life of the acid treatment and improve the outcome of the acidizing job, alternative acid blends to the conventional HCl-HF acid systems were set as part of the remedies employed in the decision trees (Fig. ).
In one of the improved chemistry systems, HCl is replaced with a phosphonic acid complex which has five available hydrogen ions that dissociate at different stoichiometric conditions. For this reason, the phosphonic acid complex is referred to as a five-hydrogen (HV) complex (Nwoke et al.; Uchendu and Nwoke; Rae and Di Lullo ). The HV acid reacts with ammonium bifluoride (NH 4HF 2) or with ammonium fluoride to produce HF acid. In order to produce a 1% HF acid solution, 20 gal of HV acid per 1,000 gal of water are required to react with approximately 123 lbm of NH 4HF 2 (Uchendu et al. This self-generating reaction of HF acid reduces the rate at which the acid system reacts, and therefore, allows an increased depth of penetration of live HF acid into the formation. This slow reaction reduces the risk of formation deconsolidation in the near-wellbore area, unlike the conventional HF treatment which may deconsolidate the formation in the near-wellbore region.In addition, the reaction of the HV:HF system with clays forms a thin aluminum silicate phosphonate coating on the clay. This coating, in return, prevents further spending of the HF acid and reduces the reaction rate on high surface area clays.
As a consequence, the volume of silica material that can be dissolved is increased, causing further improvement to the near-wellbore permeability (Obadare et al. The production of ammonium phosphonate salt, while HF is evolved from the HV acid, eliminates the generation of insoluble precipitates as the pH of the spent acid system rises. This is a common problem with conventional HF acid systems. Furthermore, the rock substrate is conditioned to be water wet with this HV:HF chemistry. Water-wet condition improves the acid contact with the targeted alumino-silicate material.
Precipitation of fluorosilicates, hexafluorosilicates, alumino-fluorosilicates, iron compounds and calcium fluoride, commonly generated during acidizing with conventional HF are prevented as a consequence of the strong chelating property of the HV:HF acid system (Nwoke et al., ). These numerous properties of the HV:HF acid system are, indeed, the reason for the improved success rate of acid jobs in several case histories.To satisfy the need for a longer-lasting stimulation effect without the generation of unwanted second-reaction and third-reaction precipitates in sensitive sandstone formations, the expert system deploys another blend referred to as the acidic chelant-based blends (Urraca and Ferenc; Rae and Di Lullo; Nasr-El-Din et al., ). The use of these acidic chelant-based blends is restricted to high temperatures formations, with relatively high carbonate content and low clay content.
The advantages of these acid blends consist of their ability to:.Dissolve both calcium and alumino-silicates.Prevent the possible precipitation of reaction by-products by sequestering many of the metal ions present in the aqueous solution: Ca 2+, Fe 2+, Al 3+ ions.Treat formations with high calcite content.Treat formations with high iron content.Treat formations with zeolite bearing minerals. System implementationImplementation of the acidizing expert system has been achieved in five phases as shown in Fig. The first phase is knowledge acquisition in which knowledge is elicited from the expert in the field. In this phase, the necessary knowledge is built up progressively through a series of consultation sessions between the domain expert and the artificial intelligence specialist, the knowledge engineer. Knowledge acquired during these sessions is recorded, refined and structured so that it could be used in the reasoning process. The main task of the second phase is to arrange the acquired knowledge in decision trees, which are considered as the main communication tools between the domain expert and the system developer.
Decision trees prepared for Stages 1–6 are shown in Figs., respectively. Main development and coding of the “Acidizing Expert System” are performed in phase three.The software used in the implementation of the system is Exsys Corvid, which is supported and licensed by Exsys ® Inc. Corvid is an expert system development tool that can be used to automate decision-making processes (Exsys Inc. Expert system development in Corvid is achieved using object structures, logic blocks, action blocks and interactive Java-based tools for Web delivery.The first step in phase three is to formulate the system by identifying and defining the variables that will be used in the reasoning process. Variables are either input variables that are acquired through interaction with the user or decision variables that are inferred and concluded by the reasoning process. The next step in phase three is to represent the knowledge, arranged as decision trees, in IF–THEN rules format. The expert system is then constructed using the development mode of Exsys Corvid.
This includes defining the variables, followed by building the questions and defining the interaction with the user. The knowledge base is then constructed by coding the IF–THEN rules taking into account the inference mechanism to be used in deriving the conclusions. During the development phase, interaction with the user is performed using Java Applet.The developed expert system is tested and validated in stage four. Ideally, testing and validation are performed by exhaustive combinations of the values of input variables.
Using the decision trees, the reasoning process is followed for each combination and the conclusions of the system are validated against the end nodes of the tree. Actual validation of the developed system is then performed by running a set of case studies that reflect practical applications.System implementation is finally concluded in stage five by delivering it to the end user. In Exsys Corvid, the developed expert system may be delivered using Java Applet or Servlet Runtime. The former is for standalone applications, while the latter is for Web-based applications. Reasoning structureThe structure of the acidizing expert system is shown in Fig. Each stage in the hierarchy corresponds to a Logic Block consisting of rule sets specialized in modeling the logic and deriving the required decisions. Moreover, each stage corresponds to a decision tree representing the knowledge elicited from the expert.
Decision trees are constructed in the previous section and illustrated in Figs.,. An acidizing consultation session starts by processing the first three stages sequentially, followed by asking about the type(s) of damage. Stages 1 and 2 are designed to treat damage type 1. Stage 3 scrutinizes conditions that wave HCl injections for preventing sludging, rigid emulsions, and iron precipitates. The reasoning process is then directed to infer the rules associated to Stage 4 and/or Stage 5, based on the selected damage type(s). Damage types 3, 4, 5, 6 and 7 are covered by the HCl pre-flush of Stage 4 (Fig. ). The reasoning process then proceeds to Stages 5 for deriving the recommendations for treatment of fines migration (damage type 2, Fig. ).
Finally, the reasoning process proceeds to Stage 6 which is concerned with the fluid selection for the post-flush (Fig. ).The reasoning protocol outlined above is defined in Exsys Corvid using a “Command Block”, which is shown in the Exsys Corvid Window capture shown in Fig. The statements listed in the command block are normally executed sequentially. After displaying the title page (TITLE statement), the system is directed to derive the value of the “Confidence” variable StageTwo. Consequently, backward chaining will be invoked, which would process the logic blocks for both Stage 1 and Stage 2. The third statement/command is RESULTS, which displays the results or conclusions reached so far.
This is followed by a command to derive the “Confidence” variable StageThree.
Summary. Exsys Corvid Core – Exsys Corvid Core is a tool for building and deploying online expert systems. It provides non-progrInfo. License$149.99. version1.0.2. File Size25.5 MB. RealeseFeb 20, 2014.
DeveloperExsys Inc. O.S.Mac OS X 10.9 or later, 64-bit processorDescription. Exsys Corvid Core is a tool for building and deploying online expert systems.
It provides non-programmers a way to easily build interactive Web applications that capture the logic and processes used to solve problems and deliver it online to any Browser – even iPhones and iPads. Organizations can now optimize their most valuable asset, expert knowledge, through powerful interactive Web-enabled knowledge automation expert systems. Online sessions emulate a conversation with a human expert asking focused questions and producing customized recommendations and advice. The decision making skills of your top experts can now be made available Screenshot.Download URL. Summary Exsys Corvid Core - Exsys Corvid Core is a tool for building and deploying online expert systems.
It provides non-progr. License$149.99 version1.0.2 File Size25.5 MB RealeseFeb 20, 2014 DeveloperExsys Inc O.S.Mac OS X 10.9 or later, 64-bit processor Description Exsys Corvid Core is a tool for building and deploying online expert systems. It provides non-programmers a way to easily build interactive Web applications that capture the logic and processes used to solve problems and deliver it online to any Browser - even iPhones and iPads. Organizations can now optimize their most valuable asset, expert knowledge, through powerful.