This section presents the proposed approach to support the design and management of modular reconfigurable pallets. The approach is based on the work by Urgo et al. [28] and consists of a set of methods and tools supporting most of the activities described in the previous section. Herein, the activities Construct, Install & Maintain System (A2) and Execute machining process (A34) are considered as out of scope. Furthermore, the assumption is made that the following activities have already been carried out: Design System Configuration (A11), Plan Processes (A12). Therefore, it is already known which are the selected production resources (i.e. machine tools, transporters, physical pallets, modular fixtures, etc.) and the set of machining operations to execute for each part type.
The following sub-sections will delve into methods and tools composing the proposed approach that is shown in Fig. 3.4. The approach resembles a star network because the need of interaction and data exchange between the tool is operated through a shared factory model based on a common Factory Data Model (component 1 in Fig. 3.4) presented in Sect. 3.4.1. The methods and tools (components 2–6 in Fig. 3.4) are described in Sects. 3.4.2–3.4.6.
3.4.1 Factory Data Model
The data model is a key element to support interoperability between different digital tools, providing the capability to retrieve, store and share information. Therefore, a suitable data model must be able to cover and integrate heterogeneous knowledge domains, while guaranteeing extensibility.
Semantic Web technologies and in particular ontologies can be employed to meet these requirements [29, 30]. Herein, a modular ontology-based Factory Data Model is proposed to formalize the information that is in particular relevant to the design and management of modular reconfigurable pallets. The OWL 2 ontology language [31] is adopted and the work of Pellegrinelli et al. [24] is taken as the basis of the proposed Factory Data Model that aims at representing a detailed pallet structure and its fixture elements, workpiece setup, pallet inspection systems, and the evolution of the state of the factory objects. The architecture of the data model is shown in Fig. 3.5 and consists of the following ontology modules:
-
list, an ontology defining the set of entities used to describe the OWL list pattern [32].
-
express, ontology mapping the concepts of EXPRESS [33] language to OWL [32].
-
IFC_ADD1, the ifcOWL ontology that is converted from the IFC standard defined in EXPRESS language [26, 32].
-
IFC_ADD1_rules, an enhancement of the ifcOWL ontology with axioms derived from WHERE rules in the original IFC EXPRESS schema [34, 35].
-
fsm, an ontology defining the concepts required for modelling finite state machines [36].
-
sosa, the Sensor, Observation, Sample, and Actuator (SOSA) Core Ontology [37].
-
ssn, the Semantic Sensor Network Ontology [37].
-
statistics, an ontology that defines probability distributions and descriptive statistics concepts [38].
-
expression, an ontology modelling algebraic and logical expressions [38].
-
osph, an ontology modelling Object States and Performance History, while integrating the ontology modules fsm, statistics, ssn, sosa, expression [38]
-
IFC_ADD1_extension, an ontology module integrating osph and IFC_ADD1_rules modules, while adding general purpose extensions to IFC_ADD1.
-
ISO14649-10, based on the STEP-NC standard [27] converted from EXPRESS schema into OWL ontology according to the pattern defined in [32].
-
factory, a specialization of IFC_ADD1 with definitions related to production processes, part types, manufacturing systems and machine tools.
-
inspection, an ontology defining the concepts of object inspection and 3D scanner system for data acquisition (e.g. laser scanner and its components).
-
FO1.3: a fragment of the FixOnt ontology presented in [23].
-
dmanufacturing, ontology module integrating four modules (ISO14649-10, factory, inspection, FO1.3), while adding specializations for the discrete manufacturing domain.
The modules IFC4_ADD1_extension, factory, inspection, and dmanufacturing represent an evolution of the modules IFC4_ADD1_extension, FactoryDomain, VisualInspectionDomain, and DiscreteManufacturingDomain, respectively, presented in [24].
The Factory Data Model can be instantiated to generate libraries of objects such as part types, operations, fixtures, and pallet components that can be later exploited to define specific pallet configurations, as discussed in the following subsections.
3.4.2 Pallet Configuration
Optimizing the pallet configuration means to identify the best location of the workpiece on the baseplates and their positioning on a given structure, taking into account the setups necessary for machining the workpiece as well as its operations (Fig. 3.6). The proposed method for pallet configuration is able to manage zero-point clamping systems with two sequential steps: accessibility analysis and workpiece allocation, thus supporting activity A13 in Fig. 3.2.
Accessibility analysis requires as input the identification of the baseplates that can be mounted on the pallet, the dimensions of each baseplate, the setups for each considered workpiece in terms of operation tool access directions and the number of axes of the machine tool. For each workpiece setup and for each baseplate, the admissible patterns (i.e. number of rows and columns of workpieces) are listed. The pattern admissibility depends on the spatial dimensions of the workpiece on the zone, the tool access directions that have to be respected for all the workpieces of the pattern as well as on the position of the face with respect to the position of the zone in the physical face of the pallet.
Workpiece allocation defines the configuration of the pallet by identifying the position and numbers of the workpieces on the baseplates and, consequently, on the pallet. Specifically, the best combination of admissible baseplates on the pallet is selected among all the possible combinations. A mixed integer mathematical model (MIP) is optimised by maximising the number of finished produced parts while meeting the constraints related to pallet balancing (the number of workpieces in each work piece setup has to be equal for each part type) and the consistency between the baseplates, the physical pallet face and the workpiece setups. The model can be re-run multiple times to generate different solutions in terms of pallet balancing and/or placement of the workpieces. More details about the pallet configuration method can be found in [12].
3.4.3 Design and Setup of the Pallet Scanner
The proposed approach performs the inspection of the pallet configuration at the shop floor level by means of a 3D laser scanner that acquires the pallet data as a point cloud. The laser scan technology is promising as a versatile and low cost solution capable to operate under difficult shop floor conditions in term of light sources and dust.
The design of the pallet scanner (i.e. part of activity A11 in Fig. 3.2) aims at supporting the complete acquisition of the pallet while considering the characteristics and size fixtures that can be used to generate pallet configurations. The following activities are carried out to design the scanner configuration and distance from the inspected object (i.e. pallet) [13, 40]: selection of the camera with optics guaranteeing the appropriate resolution; selection of laser type and optics with fan angle and positioning to cover the required volume; identification of step motors for the selection of the number of laser edges.
The point cloud generated when the scanner inspects a pallet must be elaborated and compared with the desired pallet configuration, i.e. the master geometry associated with the correct positions and shapes of all the elements composing the pallet configuration. The generation of the needed master geometries is part of activity A14 in Fig. 3.2. Two possible ways of generating the master geometries can be foreseen:
-
1.
Empirical generation. The possible and relevant pallet configurations are implemented and each face of the pallet is acquired by the laser scanner. The resulting point clouds are elaborated and the master geometries of the various pallet faces (with fixtures and workpieces) are stored in a database for future use. This solution can be managed if the number of possible pallet configurations (or at least the number of different pallet faces) is not too high; otherwise the workload of the setup phase becomes excessive.
-
2.
Model-based generation. The master geometry is calculated by exploiting the 3D CAD models of the pallet together with the topological configuration of the laser scanner. The following comparison with the scanned point cloud can be made easier by computing the area of the master geometry corresponding only to the portion of the pallet that is actually visible and acquired by the laser scanner. Therefore, it is necessary to identify all the mesh elements of the 3D CAD models that are simultaneously visible by the camera and the laser of the scanner [12]. These two sets are obtained by considering the scanner camera and the laser positions to determine the viewing frustum as in 3D computer graphics. Furthermore, critical elements that are almost parallel to the view direction can be removed to reduce the noise during the comparison elaborations.
The proposed approach is able to deal with both ways of generating the master geometries, therefore it is not necessary to impose restrictions.
3.4.4 Loading and Optimization
As anticipated in the introduction, the use of zero-point fixture systems can actually increase the flexibility of an FMS. In particular, the possibility to change the fixtures mounted on a pallet in a short time can lead to:
-
Lower number of pallets and fixtures needed to satisfy the same demand mix.
-
Better workload optimization for all the involved resources (machine tools, pallets and fixtures).
-
Shorter makespan to accomplish the overall required production.
These options open new ways for production planning policies never considered in literature because not applicable with traditional clamping technologies. Therefore, herein the aim is to support activity A31 (see Fig. 3.3) by optimizing the reconfiguration of physical pallets (i.e. changing the assignment of the available baseplates to the tombstones) and the assignments of machining operations to machine tools. The set of machining operations associated with a pallet configuration can be assigned to more than one machine tool thanks to the non-linear formulation of the part program [41, 42], thus increasing the complexity of the loading problem.
The proposed Loading and Optimization method jointly tackles the fixture assignment and loading problem by sequentially solving two sub-models (see Fig. 3.7) to reduce the computational complexity while introducing a certain degree of approximation:
-
1.
Model 1 assigns the fixtures (i.e. baseplates) to the pallets over the planning horizon, thus determining the needed pallet reconfigurations. The goal is to minimize both the required number of baseplates and the makespan.
-
2.
Model 2, given the outcome of the first model, assigns the machining operations of the baseplates to the available machines (or groups of machines), thus determining the routing of the pallets in the system. The goal is to minimize the makespan while taking into consideration a higher level of details with respect to Model 1.
The proposed method has been tested on 10 production problem instances described in Table 3.1. The results of the proposed method have been compared with the results that can be obtained without zero-point clamping technology, i.e. using traditional pallets. The same number of physical pallets has been considered in both cases, therefore the cost for the pallets and fixturing equipment is equivalent, except the additional cost for the baseplates endowed with zero-point clamping technology.
Table 3.1 Production problem instances
The experiments showed a saving of about 10% on average in terms of makespan thanks to the use of zero-point clamping technology compared with the use of traditional pallets. This means that, if the benefit linked to this time saving is higher than the cost of adopting the zero-point clamping technology, then it is advised to opt for it.
3.4.5 Pallet Inspection
The activity Make Pallet Inspection (A32 in Fig. 3.3) is carried out by scanning the actual pallet configuration that must be processed by the machining centres. The pallet scanning returns a point cloud that must be further elaborated to check if the pallet configuration is correct. This check is performed through the evaluation of the difference between the scanned point cloud (i.e. slave geometry) and the master geometry (Fig. 3.8) generated during the setup of the scanner (see Sect. 3.4.3).
Each point of the slave geometry is analysed and the three closest points in the master geometry are selected to calculate a plane equation. Then, the distance between the point of the slave and the plane is computed. Finally, the minimum square error based on all the computed distances is calculated. After the elaboration, the method returns statistics in terms of maximum, mean, and minimum errors with respect to the reference pallet configuration [12]. These results can be used to support a decision system identifying significant deviations as a symptom of possible errors in the mounting of the parts or be presented to the operator who decides if any intervention is needed.
3.4.6 Machining Process Simulation
The activity Generate & Validate machining process (A33 in Fig. 3.3) consumes the results of the pallet inspection together with the information related to the machining operations and the fixture and pallet configuration. The part program associated with the whole pallet can be automatically generated while adopting the Network Part Program method mentioned in Sect. 3.2. The part program is defined in terms of a partially ordered set of operations exploiting the STEP-NC [27] structure through machining working steps (MWs). A set of MWs is associated with each baseplate, whose toolpaths are referred to the coordinates of the baseplate itself. Once the baseplates are located onto the tombstone and recognized through the pallet inspection, the coordinates of the baseplates are updated accordingly and the actual paths for the tool are derived automatically. Specific algorithms are used to generate the rapid movements between the baseplates taking into consideration their placement as well as the geometry of the fixture, in order to avoid collisions. Advanced algorithms can be employed to optimize the process for the whole system, i.e., clustering operations sharing the same tool to minimize the tool changes.
The commercial software tool VericutFootnote 1 has been adopted to simulate the part program and validate the CNC machining while considering the static and dynamic properties of the machine tools. This process simulation is able to check if the machine tool can perform the required machining operations without collisions.