In today's global economy, manufacturing enterprises must employ increasingly effective and efficient information systems. Such systems should result in the seamless integration of manufacturing applications and exchange of manufacturing process information between applications. The Industrial Ontologies Foundry (IOF) seeks to create a set of core ontologies that spans the entire domain of digital manufacturing. In this paper we explore the adequacy for the Process Specification Language (PSL) to serve as the foundational ontology that can be used to axiomatize the ontologies within IOF. We provide conservative definitions for the Top 20 Classes within the IOF signature, and show how PSL formalizes the competency questions for the IOF ontologies.
The goal of the Industrial Ontologies Foundry1 is the creation of a set of core ontologies that spans the entire domain of digital manufacturing. This set of ontologies will be non-proprietary and are expected to serve as the foundation from which other domaindependent and/or application ontologies can be derived in modular fashion across all industrial domains and manufacturing specializations. The scope spans all industrial domains and types of manufacturing (batch/continuous process, discrete high/low volume, make-to-order), all operational areas of a manufacturing enterprise (design, engineering, manufacturing, sourcing, supply-chain management), and all stages of the product life cycle from inception through end of life.
One approach to the design of the IOF Ontologies is to use a foundational ontology
as the basis for the axiomatization of the intended semantics of the classes, relations, and
functions in the signature of the IOF Ontologies. The Process Specification Language
(PSL) ([
... .. . ..
6
6 . ....
Ta.tomic . . .
Tduration 6 .. ... .....
... ..
Tdisc state 6 . .. ..
.... ..
The Process Specification Language (PSL) has been designed to facilitate correct and complete exchange of process information among manufacturing systems Included in these applications are scheduling, process modelling, process planning, production planning, simulation, project management, workflow, and business process reengineering.
PSL is a modular ontology, with each module consisting of at most twenty axioms. Overall, there are approximately 250 axioms spread across 31 modules. (This counts only those CLIF texts that axiomatize primitive concepts, since we have adopted the practice in which defined relation is found in its own CLIF text). The modules of the PSL Ontology2 can be seen in Figure 1.
The goal for the TUpper Ontology is to support the ontological analysis of relevant
existing standards and to integrate the ontologies within those standards ([
boxworld cardworld component matter_change motion multidim_motion
Morpheus
assembly codi
OVid
FOUnt ProSPerO of the TUpper Ontology is determined by the task of standards integration – each of the generic ontologies is needed to support one of the participating standards. Each of the generic ontologies is a reductive module of TUpper. There are no additional bridge axioms in a combined signature across these modules. The verification of TUpper is therefore trivial, being equivalent to the verification of each of the generic ontologies.
TUpper also contains modules for reasoning about state designed using the
methodology of [
As proposed in ISO 21838-4, TUpper (which includes PSL as a module) is a top level ontology since it satisfies the criteria for breadth of coverage.
Space and Time As axiomatized by Tpslcore, an object exists at a timepoint and an activity occurs at a timepoint. Each activity occurrence has a beginof timepoint at which it starts to occur and an endof timepoint at which it ceases to occur. Similarly, each object has a beginof timepoint at which it starts to exist and an endof timepoint at which it ceases to exist. Timepoints are linearly ordered by the before relation, with endpoints at infinity. The existence of timeintervals is axiomatized in Tinterval psl – each interval corresponds to a pair of timepoints.
Tregion mt is the mereotopology that is specified over the set of spatial regions. Toccupy root is the module that is the location ontology in TUpper. It is consistent with the ontology for more than one material object to occupy exactly the same spatial region at the same time.
Actuality and Possibility Models of Tocctree are occurrence trees, which consist of all possible sequences of atomic activity occurrences. The set of activities that actually occur in a model are elements of one branch of the occurrence tree. Models of Tcomplex consist of subtrees of the occurrence tree that correspond to possible occurrences of complex activities. A legal occurrence tree is the subtree of an occurrence tree in which all activity occurrences satisfy precondition axioms that specify the conditions under which an activity can possibly occur. Dispositions are treated via such precondition axioms. Parts, Wholes, Unity, and Boundaries TUpper adopts mereotopological pluralism – there are multiple distinct parthood and connection relations for different classes of entities. The mereologies on matter and spatial regions are complete extensional mereologies with complementation, as is the mereology of atomic (concurrent) activities. On the other hand, the mereology of complex activities is logically synonymous with the weakest mereology, and does not require the existence of sums or complements. The mereology on components and timeintervals entails Strong Supplementation, but does not require the existence of sums for underlapping parts.
Space and Place Toccupy is the module that is the location ontology in TUpper. Tmotion classifies all activities that can possibly change the location of an object.
Shape is represented using the Tboxworld module of TUpper. Holes are specified as physical objects which have shape but which are not constituted by matter.
While Tmatter [ Tconstitution axiomatizes material objects and their constitution by matter, Tmatter change classifies all activities that can change the mereology of matter and that can change a material object by changing the matter that constitutes the object. Quantities and Mathematical Entities The FOUnt module within TUpper axiomatizes the units of measure, constraints on how units can be algebraically manipulated), and the relationship between physical objects and processes and the units of measure. Processes and Events Processes appear as the class of activities within Tpslcore. Within the module Tdisc state, fluents (states) are achieved or falsified by activity occurrences, and only activity occurrences can change fluents. There are no constraints on the kinds of processes that exist. Each activity occurrence has a duration (Tduration), which can be zero in the case of an instantaneous activity occurrence (i.e. one in which the beginof timepoint is equal to the endof timepoint).
Constitution Tconstitution axiomatizes the constitutes relation between an object and its matter. Constitution only holds between objects and matter, and there is no analogue of constitution that holds between processes or nonmaterial entities.
Causality Constraints are definable between activities, as well as between activities and fluents (states). In particular, one activity can achieve the preconditions of another activity, or a fluent may trigger the occurrence of an activity.
The PSL Ontology is verified – the models of all core theories within the ontology have
been characterized up to isomorphism ([
Tbounded linear order [ Tscatter graph partitioning [ Tgraphical incidence [ Tpresent bundle The consistency of PSL-Core follows from the consistency of each of these mathematical theories.
TUpper and PSL have a natural language representation, and they are axiomatized in both Common Logic and OWL-2. However, the OWL-2 is far too weak to satisfy any of the initial requirements or to support any of the motivating scenarios in decision support and semantic integration in the next section of this paper. In particular, any OWL-2 axiomatization will have unintended models, and the existence of such models prevents the specification of reasoning problems for planning and scheduling. It is therefore not clear how the OWL-2 axiomatization of PSL would be used.
Many tasks require correct and meaningful communication and integration among intelligent agents and information resources. A major barrier to such interoperability is semantic heterogeneity: different applications, databases, and agents may ascribe disparate meanings to the same terms or use distinct terms to convey the same meaning. The development of ontologies has been proposed as a key technology to support semantic integration—two software systems can be semantically integrated through a shared understanding of the terminology in their respective ontologies.
A semantics-preserving exchange of information between two software applications requires mappings between logically equivalent concepts in the ontology of each application. The challenge of semantic integration is therefore equivalent to the problem of generating such mappings, determining that they are correct, and providing a vehicle for executing the mappings, thus translating terms from one ontology into another.
The work by [
SAP ERP ([
MetCapp ([
ILOG Scheduler ([
A reproduction of this use case with the other foundational ontologies (BFO and DOLCE) for the IOF Ontologies would be another approach to the evaluation of the fitness of the foundational ontologies.
Whereas use cases for semantic integration are based on existing engineering software applications, the IOF Ontologies are also intended to support the design of new manufacturing software systems that automate decision support tasks performed by engineers at all phases of the product lifecycle.
Scenario 1: Manufacturability As products are being designed, software automatically checks to determine whether the product will be manufacturable, that is, whether there exists a process plan whose execution produces objects that satisfy the product’s quality constraints. If a design is not manufacturable, potential modifications to the design are recommended.
Scenario 2: Proactive Manufacturing A new approach to achieve this integration has
been the notion of proactive computing, in which information systems act in anticipation
of future problems, needs, or changes of the user. To be proactive, a computer system
must understand the user’s context and how it changes over time. Within
manufacturing enterprises, this can be facilitated by the recording of process constraints associated
with a particular resource as it passes through the set of activities performed within the
supply chain. RFID technology can be combined with ontologies (e.g. use a process and
time ontology to store information directly on the RFID tags) to create smart items and
demonstrate this approach using a motivating scenario of manufacturing process control
[
Several automated reasoning problems arise within the motivating scenarios introduced in the preceding section. In this section, we give a brief overview of how PSL can be used to represent these problems so that they can be treated as competency questions relevant for the IOF Ontologies.
The problem of temporal projection is closely related to prediction about how the world changes: given a set of activity occurrences and an initial state, what is the final state?
state SF , determine whether the ground state formula Q(O) holds in after the activity occurrence O:
Tpsl [ Socc [ SF j= Q(O)
It is important to understand the significance of this formal definition of temporal projection. Given a sequence of activity occurrences, we must determine from the axioms in Tpsl alone whether the formula Q(O) is entailed after the activity occurrence O. We therefore need to determine the necessary set of axioms in such a process ontology in order to specify the problem of temporal projection and to characterize the solutions to this problem.
What are the minimal ontological commitments to define temporal projection? Since we are given a sequence of activity occurrences, and we must determine what fluents (states) hold after the sequence, we need to axiomatize preconditions (fluents that must hold in order to perform an action) and effects (fluents that hold after performing an action).
The problem of plan existences asks whether there exists a sequence of activity occurrences such that a given state holds after the sequence. Such a set of activity occurrences constitutes a plan that achieves the goal state.
Problem 2. Given a simple state formula Q(o), an initial state SF and a set of precondition and effect axioms Sactivity determine
Tpsl [ SF [ Sactivity j= (9s) Q(s)
Given a set of deadlines, operations, and resources, does there exist a sequence of activity occurrences that achieve a set of goals by the deadlines?
ground term T , determine
Tsucc [ Tpre [ SF j= (9s; t) S0
s ^ Q(s) ^ start(s; t) ^ t < T
This is equivalent to the existence of a plan that can achieve the goal Q(s) by the deadline T .
Process plan generation differs from plan existence in two ways. First, the process plan is a complex activity (that is, an activity composed of primitive activities with partial ordering constraints and nondeterminism). Second, the intended effects of occurrences of the process plan must be equivalent to the specification of a product. Process plan generation is the construction of a plan whose occurrences produce an object with a given set of features.
Problem 4. Given a product specification formula Sproduct (x) for a product P(x), find a process description Sactivity(A) such that Tpsl [ Sactivity(A) [ Sproduct (x) j= (8o; x) occurrence o f (o; A) ^ P(x) holds(product(x); o)
Given a partially order complex activity that specifies a plan, is the goal achieved after every occurrence of the plan?
determine
Tpsl [ Sactivity(A) j= (8o) occurrence o f (o; A) Q(o)
Closely related to plan verification is the problem of reasoning about the interaction of a plan with external activities (i.e. activities that are performed by actors outside of the plan). Which activities external to the plan can interfere with the plan? Must activities external to the plan occur in order for the plan to occur?
The set of terms in the initial IOF signature focus on classes of resources and processes within manufacturing. In this section, we present the axiomatization of the intended semantics of the IOF signature using PSL, as required to support the competency questions specified in the preceding section. Note that the axioms for this initial set of terms in the IOF Ontologies forms a definitional extension of the PSL Ontology; no additional ontologies outside of PSL and TUpper were used.
Several classes within the IOF Top 20 terms are related to objects that are resources for manufacturing processes. The challenge is to provide an axiomatization that distinguishes resources from other objects that participate in an activity occurrence. Within the PSL Ontology, resources exist wherever there are concurrency constraints on activities – if two activities cannot be concurrent, there exists a resource that they both require. The different relationships between resources and activities are classified by the roles that the resources play in the activities.
A resource r is consumable by an activity a if any other activity that also requires r is not possible to perform after a completes its occurrence. Examples of consumable resources include wood in a fire, or raw materials in a manufacturing production process.
(8x) iof:MaterialResource(x) (9a) MaterialOb ject(x) ^ consumable(a; x) (1)
A resource r is reusable by an activity a if any other activity that also requires r is still possible to perform after a completes its occurrence, in every possible future. A hammer or screwdriver is an example of a reusable resource – as soon as one activity occurs, it is always possible to perform the next activity.
A resource r is weakly reusable by an activity a iff for any other activity that also requires r is still possible to perform after a completes its occurrence, in every possible future situation unless it is prevented. For example, a paintbrush is reusable only if we put it into varsol after use; otherwise, it is not reusable.
activity.
(8x) iof:ManufacturingTool(x) (9a) solid physical ob ject(x) ^ reusable(a; x) (2)
A resource r is possibly reusable by an activity a iff for any other activity that also requires r is still possible to perform after a completes its occurrence, in some possible future situation. A machine that requires some setup between different activities. After the first activity occurs, it is possible for the other activity, but only if the setup activity occurs first.
reusable by an activity.
(8x) iof:ManufacturingMachine(x) (9a) solid physical ob ject(x) ^ possibly reusable(a; x) (3)
A resource r is wearable with respect to an activity a1 iff after the occurrence of a1 there is always a situation in every possible future where any other activity that requires r is no longer possible. An example of a wearable resource is a drill bit – in every possible future, there will exist a situation where the bit has worn down to the point where it can no longer be used.
(8x) iof:Fixture(x) (9a) solid physical ob ject(x) ^ wearable(a; x)
The PSL Ontology axiomatizes other classes of resources that play a role within planning and scheduling, although they did appear in any of the IOF Top 20 Terms.
A resource r is weakly consumable with respect to an activity a1 iff after the occurrence of a1, there always exists a possible future along which any other activity that requires r will never be possible. Consider, for example, a paintbrush – if we do put it into varsol after using it, then any activity that requires the brush will no longer be possible.
A resource r is renewable with respect to an activity a iff for any other activity that also requires r is still possible to perform after a completes its occurrence, in every possible future situation unless it is prevented. An example of a renewable resource is a solar-charged battery – once it is depleted, there will always exist a future situation where the sun recharges the battery so that it can be used again.
PrOSPerO (Process Ontologies for Solid Physical Objects) is the collection of domain process ontologies within TUpper that axiomatize the classes of activities that change objects within manufacturing.
The approach [
With ProSPerO, we treat SoPhOs and Occupy as the domain ontologies, leading to the following domain process ontologies:
Material removal / addition activities Shape-changing activities Assembly / disassembly activities
Motion activities
In particular, there is a domain ontology that axiomatizes the mereological relation for the components of a solid physical object.
Definition 6.
(8x) iof:Assembly(x) solid physical ob ject(x) ^ (9y) component o f (y; x) ^ (x 6= y)
(8x) iof:Component(x) solid physical ob ject(x) ^ ((8y) component o f (x; y) (x = y)) (5) (6)
Within the corresponding domain process ontology, assembly and disassembly activities change the comp(x; y) mereological fluent.
(8x) iof:AssemblyProcess(x) activity(x) ^ ((8o) occurrence o f (o; x) (9y; z) achieves(o; comp(y; z)) Similarly, transportation processes change the location of objects.
(8x) iof:TransportationProcess(x) activity(x) ^ ((8o) occurrence o f (o; x) (9y; z) achieves(o; loc(y; z))
Definition 9. 5.9. Processes: Composition (7) (8) (9) (10) (11) (12) Activities can be composed together to construct complex activities, or decomposed into primitive activities. Occurrences of complex activities correspond to sets of occurrences of their subactivities. Different occurrences of complex activities may contain occurrences of different subactivities or different orderings on the same subactivity occurrences. There is a partial ordering of subactivity occurrences for a set of complex activity occurrences.
5.10. Plans (8x) iof:Task(x) primitive(x) A processor activity is an activity which uses some set of resources, consumes or modifies some other set of resources, and produces or modifies a set of objects. (8x) iof:ManufacturingProcess(x) processor activity(x) A set of activity occurrences is a plan in the PSL Ontology if and only if all occurrences of subactivities that agree on state also agree on being subactivity occurrences. (i.e. there exists a set of fluents that are achieved by all occurrences of the activity).
(8a) iof:Plan(a) ((8o) root(o; a) plan(o))
(8x; o) P(x) (prior(product(x); o)
F(x; o) where P(x) is a specific class of objects and F(x; o) is a formula.
(8a) iof:Process plan(x) (13) ((8o) iof:plan(a) ^ occurrence o f (o; a) (9x) achieves(o; product(x))
An activity occurrence occ2 is the next processor subactivity occurrence after occ1 in an activity a iff the output material of a1 is the input material of a2, and there is no other processor subactivity of a which consumes the output material from a1, and which occurs between a1 and a2. There exists a material flow ordering, which is a partial ordering over the processor subactivity occurrences of a with respect to resource flow. An activity is a resource path iff the subactivity occurrence ordering is equivalent to the material flow ordering.
plan for a product.
(8a) iof:ManufacturingProcessPlan(a) (9x) iof:Process plan(a; x) ^ resource path(a) (14)
We have seen how the PSL Ontology, together with the modules of the TUpper
Ontology, can be used to provide conservative definitions that axiomatize the intended
semantics for the Top 20 Classes in the IOF signature. Furthermore, PSL has the
expressiveness to formally specify the competency questions that arise from motivating scenarios
within IOF, and earlier work (e.g. [
Nevertheless, major challenges remain unresolved. What role does the foundational
ontology play? Can multiple foundational ontologies be used together for the
axiomatization of the IOF Ontologies? For example, can the mappings between PSL and DOLCE
[
More importantly, we still do not know what we need to axiomatize – the IOF community has not yet agreed upon the signature of the IOF Ontologies, let alone the intended semantics of the signature. Without a specification of the intended semantics of the IOF signature, it is difficult to make any claims about the adequacy of an existing ontology to be reused to axiomatize the intended semantics.
Finally, many of the issues that are being encountered in the design of IOF are open research problems in ontological engineering. We must be careful to avoid easy approaches that do not ultimately lead to solutions to these problems.