【正文】 unique. The Type is the keyword of the feature. Thefeature Location is about the original position anddirectional vector of the feature. The feature size, . including final size FinalSize and initial size Init.Size , is the dimensional value of the feature. TheHardness is about the feature’s hardness. The Tolerances are about dimensional and geometric tolerances. A keyword is assigned to tolerance to distinguish the tolerance content.. Domain knowledgeThe domain knowledge of each Pagent is aboutthe descriptions of its process planning capabilities,and used to generate proposals, evaluate proposals,and resolve conflicts. Different agents may havedifferent formats for domain knowledge representations such as databases, analytical algorithms, etc.The domain knowledge can be extracted from manuwxfacturing handbooks such as 19–21 . Each Pagenthas a different domain knowledge content.For example, the operation selection agent is usedto generate machining operation alternatives for defined features of parts. For each given feature, theremay exist more than one possible operation. Sometraditional machining methods such as forging, diecasting, drilling, turning, milling, boring, shaping,grinding, lapping, honing, and diamond boring, etc.,have been built into the domain knowledge base. Itscontent includes the relationship between operationswith parameters such as machinable feature, workpiece material, tolerances, preparatory operation, timecalculation equation. The knowledge is representedas facts in a semantic as shown in Fig. 3. In theknowledge base, an operation fact is stored in onenode. All facts are stored in a list.The proposal generation, proposal evaluation, andconflict resolution strategies are embedded in thesolver adapter of each Pagent.Fig. 3. Semantic for internal round features.. Conflict resolution knowledgeThe conflict resolution strategies include two categories: domaindependent and domainindependent.The domaindependent strategies mainly involve howto suggest further measures when one Pagent conflicts with the other Pagent’s proposals. Each Pagenthas its own suggestion strategies different from theother Pagent’s. It also includes the explanation tothe conflict resolution proposed. The domainindependent strategies are more mon and can be thesame for all the Pagents. They are designed as a setof conflict resolution facts with some basic guide .lines conflict resolution rules for deciding whichdomaindependent strategy to apply and altering theresolution strategy in order to improve its understanding of the overall problem according to theother Pagent’s action of conflict resolutions. Thefollowing strategies are used to resolve conflicts inthe CoCAPP system as domainindependent conflictwxresolution strategies 22 .fl Compromise: finding an immediate proposal thatis within an acceptable range。fl CasebasedParameterSetRetrie?al: finding a previous solution that succeeded in resolving a conflict involving a similar set of parameters.. Control knowledgeAll the Pagents in the CoCAPP system must beable to perform a set of tasks with the same controlknowledge:fl generate new proposals。fl negotiate。fl internal knowledge representations which capturesufficient goal and history information to allowfor solution revision to be carried out cooperatively。fl mechanisms for incorporating externally produced partial solutions。fl the ability to coordinate an internal agenda withexternal events.Each Pagent municates with other agents byusing a mon shared language. The proposal generation and evaluation, solution, and conflict generation and resolution are produced according to theinternal domain knowledge of the Pagent.4. Software implementationwxThe IBM ABE Toolkit 24 is chosen as thedevelopment environment of the CoCAPP system.The Visual Cqq language is chosen as the implementation language. The system can run in the platform of Windows 95 or Windows NT.Fig. 5. Conflict resolution of Pagents.(). Zhao et in Industry 41 2020 83–9792As mentioned, each Pagent in the CoCAPP system makes use of three types of knowledge: domainknowledge, control knowledge, and conflict resolution knowledge. In order to enhance the flexibilityand scalability, the domain knowledge is furtherclassified into universallevel, shoplevel, and machinelevel knowledge. The universallevel knowledge is applicable to any status without consideringindividual panies and is established when theCoCAPP system is in development and is often fixedafter the system has been constructed. The shopleveland machinelevel knowledge can be added andmodified by individual panies when the systemis scaled. The machinelevel knowledge is only applicable to a specific machine. The knowledge ofeach agent is implemented as a database or file.The conflict resolution handler is shown in Fig. 5.It consists of two effectors: ‘WatchConflict’ and‘ResolveConflict’, as well as one conflict resolutionfacts file and one conflict resolution rules file. The‘WatchConflict’ is used to map out the conflictproblem space and conflict situation. The ‘ResolveConflict’ is used to resolve conflicts. The conflictresolution strategies are implemented in this effector.Fig. 6. Test ponent.5. Case studyThe case study is used to illustrate the characteristics of the Machining CoCAPP system. The firstexample demonstrates the feedback due to an unreasonable product design. In this example, as an unreasonable part design is provided, the CoCAPP systemgenerates an ‘unresolved conflict’ output and reportsto the Dagent the locations and causes of the conflict. The part is shown in Fig. 6, a bar with anenvelope size of 50=40=30 mm3.. Initial Part Design Datatimeless:7200