Towards the Automated Coverlay Assembly in FPCB Manufacturing: Concept and Preliminary Tests

The requirements of the assembly procedure are determined based on the manual process currently carried out and the material properties.

In the manipulation of flat surfaces characterized by extremely low thickness combined with low stiffness materials, thus, deformable, the issue of preserving component configuration, avoiding undesirable bending, arises. This aspect can be even more relevant during the positioning phase, in which folds and air bubbles must be avoided. The use of a flat gripping surface as interface with the component represents a valid option to address these aspects.

Gripping strategy shall also comply with the requirement of the selective pick. Compliant grippers usually rely on the 3D shape of the object, in order to passively adapt to perform gripping, using different approaches and often implementing soft materials [8]; there are also commercial products available enabling the unconditional pick of objects with flat surfaces [9]. In the application dealt with in this paper, coverlays lie on a flat surface; due to the extremely low thickness, their 3D shape is not appropriate to be passively distinguished from the background by a compliant gripper. Thus, the necessity of a strategy for gripper “pre-shaping” arises.

Basing on these considerations, the gripper specifically designed for coverlay “pick-peel-place” is characterized by a flat interface surface and based on vacuum principle; a pattern of holes, representing themselves a matrix of suction cups, is realized in the flat surface. In order to avoid the uncontrolled pick of other patches lying on the same support, as well as the possible vacuum loss due to uncovered condition of one or more orifices, each suction cup is controlled individually, thanks to an upstream pneumatic valve module. With this approach, gripping pattern can be set depending on each specific coverlay contour, enabling the possibility of gripping different geometries.

Peel strength represents one of the most important characteristics of coverlays, as it is a measure of component integrity in determined loading conditions. DuPont^TM Pyralux^® LF coverlay material, for example, based on a Kapton^® polyimide film, exhibits a peel strength of 1.6 kg/cm after curing, defined by the provider as per the applicable standard [10].

On the other hand, peeling represents also one phase of the assembly, as the coverlay is provided coupled with a release material; it is also the most demanding challenge to address in order to accomplish the whole task, due to the uncertainties in the peeling dynamics combined with the geometrical variability of the patches. Peeling dynamics has been widely investigated by the scientific community, and there are a number of patented machines, designed to be implemented in roll-to-roll production for different applications, performing tape peeling. In the majority of cases, peeling is triggered by the action of rolls [11] or insertion means acting as blades [12], combined with high bending angles of the release material. However, circuit board industry is mostly characterized by panel-based processing and coverlays may have very different shapes and dimensions. Therefore, the necessity of implementing a specific material removal method arises, studied for components characterized by planar dimensions in the order of few square centimeters.

In order to trigger the peeling of a film, adhering to another due to a substrate of a bonding agent, a mechanical action should be applied. With decreasing thickness, this task becomes more demanding, requiring precision and tools with appropriate edges. The detachment is achieved when the force, properly applied, is sufficient to trigger local fracture of the bonding agent. In the approach hereafter presented, the issue is addressed by an inverted perspective: instead of applying a detaching force when precise positioning occurs, the method relies on driving the film towards a source of mechanical action, working continuously at a high frequency; the approaching is stopped when detachment takes place due to the contact randomly occurred between the film and the source of the force.

In the practical implementation, a rotary tool with appropriate cutting profile can be suitable. In our experiments, a rotary sanding drum was used as a trigger for the peeling of the release material. As shown in Fig. 6, such a method, called hereafter “pre-peeling”, exploits abrasive surface roughness to collide with the release material surface and detach it from the coverlay, thanks to the impulsive force obtained by high grinding speed and applied by one or more abrasive particles to the film to remove. Relying also on friction contact, mutual positions can be calibrated in order to guarantee that this force affects only the protective film. A specific setup, further described in this paper, was used to validate this method.

In Fig. 6, the most important process parameters are indicated. The a parameter is the unsupported overhanging portion of the coverlay (cantilever configuration) out of the clamp jaws, while e is the distance, along the x-axis, of the coverlay tip from the grinding tool axis, r is the radius of the sanding drum, rotating with angular speed ω and exerting a torque “T”. Parameter e is straightly related to the approach angle ϑ to the cylindrical grinding tool, since e = r cos(ϑ). These parameters have been chosen due to their direct relation with the physics behind the peeling process. Indeed, a is related to the coverlay flexural stiffness, so that stiffness decreases as a increases. Parameter e has influence on the peeling mechanism as it directly affects the direction of the interaction force applied by the tool on the component, due to the direct relation with angle ϑ. The combination of r and T determines the value of the force and ω affects the momentum. Finally, the whole process is closely linked to the roughness of the sanding drum.

Parameter h represents the distance, along the z-axis, between the sample and the grinding tool. The optimal set [a, ϑ, h] uniquely identifies the most suitable configuration of the coverlay and the rotating sanding drum for the pre-peeling phase. Regarding this aspect, it is relevant to point out that, in the practical implementation, coverlay samples preserve the curvature of the original material reel. The real h values differ from the theoretical value h* = r sin(ϑ) as they are affected by coverlay intrinsic curvature. According to this, ϑ is to be considered as the angle spanned between the radius connecting the center of the sanding drum with the contact point P and the axis of the clamp (or, in the automated task, of the end-effector), rather than the surface of the coverlay itself, as shown in the detail in Fig. 6-b).

The assembly procedure is then defined in accordance with the chosen gripping and peeling strategy. With reference to Fig. 7, the following phases are identified:

The workcell architecture for coverlay application, outlined in Fig. 8, is aimed at exploiting the proposed automation strategy to obtain the PCB assembly ready for lamination and curing. The inputs of the process are the flex or rigid-flex PCB with circuits and SMT (surface mounted technology) components installed in previous steps.

In the assembly concept, the task is customized off-line, depending on the coverlay contour: a specific software tool is required for the pre-shaping phase. This setup is fundamental for task completion, since an appropriate portion of the picked coverlay has to remain outside of the gripper envelope, enabling peeling triggering.

In the workcell concept there are two conveyors, one providing the laser-trimmed coverlays and one sliding the flex or rigid-flex PCBs to assemble. The “pick-peel-place” task is performed by a commercial manipulator, equipped with the specifically designed gripper. The implementation of a collaborative manipulator is foreseen, due to the advantages provided by including a human operator in the assembly environment [13], even for a mere on-site supervision. Within the collaborative robot scenario, several commercial models are suitable for the task, considering the extremely low payloads and the accuracy required, assessed 100 µm in Sect. 2.1. As an instance, in Figs. 8 and 9 the gripper is equipped on a Universal Robot UR5e. Regarding component pick and place, two vision systems are placed on top of the two conveyors, with appropriate fields of view (FOVs) to detect, respectively, position and orientation of the coverlay to pick, and the release position on the PCB. Within the robot reach, the pre-peeling and peeling stations are included, corresponding respectively to a rotating sanding drum and a fixed vacuum gripper; a vision system is necessary for gripping position check.

A specific gripper is designed in order to perform coverlay film manipulation and comply with all the assembly phases. The overall design is shown in Fig. 9 and each subsystems is hereafter described.

As stated above, the peeling of the coverlay is conceived in two steps: pre-peeling and peeling completion. In order to verify the effectiveness of the proposed method to trigger coverlay peeling, a series of experimental tests was performed. The following parameters are considered to mainly affect the process: thickness of the coverlay, grit designation and dimensions of the sanding drum, rotational speed, geometrical configuration of the coverlay/drum approach. In this phase, the feasibility is addressed, focusing on the individuation of the optimal geometrical configuration for a defined set of coverlay material (thickness), pre-peeling tool (grit designation and diameter) and rotational speed.

Coverlay material is DuPont Pyralux LF0110 made of 25 μm thick Kapton polyimide coated on one side with 25 μm proprietary modified acrylic adhesive. The adhesive side is covered by a protective film with thickness of 40 μm. Coverlays are IPC certified by the supplier. Coverlay geometries are determined on the basis of an industrial case, currently object of study of the research group; such a geometry is rectangular with dimensions 25 × 40 mm; further materials and dimensions will be investigated in following experimental surveys. The samples are trimmed, which means full-thickness cut, by a 2D laser cut machine from a coverlay sheet with dimensions 610 × 454 mm, obtained by cutting a roll 610 mm wide and 76 meters long.

The laser trimming operation separates the coverlays from the sheet, then collected and loaded on a conveyor for the subsequent operation. The testing equipment is composed by two main parts: 1) apparatus for coverlay positioning; 2) pre-peeling stage. With regard to the first component, the coverlays samples are mounted on a three-axes high precision metric stage, actuated by three micrometer screws, with travel range of 15 mm on each axis and sensitivity 1 μm. The coverlay samples are held inside two small aluminum jaws mounted at the end of two cylindrical fixtures. In order to distribute the clamping force on a higher coverlay surface, two stainless steel plates with thickness 0.45 mm are used as interfaces between the clamp jaws and the coverlays. The pre-peeling stage consists of an electrical drilling tool, mounted horizontally. The sanding drum is a cylinder with an external diameter of 8.78 mm and height 12.7 mm (1/2 in.) used for smoothing applications, with grit designation P80, corresponding to average particle diameter of 201 µm. The tool shaft has a diameter of 3.2 mm. The setup is showed in Fig. 10.

Preliminarily, the geometry of the set-up is verified: vertical stroke of the coverlay, alignments between grinding tool axis and coverlay profile, zeroing of the z-axis coincident to the first contact between the coverlay protective layer surface (bottom) and the side of the cylindrical grinding tool, centered position of the grinding tool with the coverlay width, etc.

The series of experimental tests was based on the variation of parameters a and e; ω was set at 800 rpm and the same sanding drum - thus, same radius and same sanding grade - was used. a was set at 10 and 15 mm, while e was set at 0, 1, and 2 mm; for a = 10 mm also e = 1.5 mm was considered as a further set of variables. Each of the 7 parameter sets is tested with 5 repetitions (sample labels A, …, E). This value is reputed sufficient for the feasibility assessment. Each run is performed by setting the parameters (a and e) and driving the coverlay sample towards the grinding tool by slow vertical motion (approximately 2 mm/min), as schematically shown in Fig. 6; vertical motion is stopped when peeling is triggered. The outputs of each experiment are the peeling triggering success, the peeled area and the recorded value of the parameter h; the standard deviation characterizing this value for each set of runs is indeed used to assess the repeatability of the process. For each successful pre-peeling test, an image was captured and processed in order to calculate the peeled area, considered an important parameter, since peeling completion relies on the gripping of a film portion already peeled.

Figure 11 shows the results of image processing procedure previously described, obtained for two runs. The light brown portion of images are pre-peeled areas (Fig. 11, left). They have irregular contours which are identified by image processing algorithms (Fig. 11, right). The proposed procedure can also be automated and used real-time for peeled area detection and characterization. Accuracy can be increased by illumination and background optimization in the image acquisition step.

Table 1 reports the results obtained with the different parameter sets. At a first glance, it highlights that the grinding-based method can be a successful solution for peeling triggering in several configurations. It is important to point out that all the successful pre-peeling runs led to successful peeling completion, definitely validating the approach of splitting the peeling issue in two sub-phases.

Besides the validation of the approach, this preliminary series of tests was aimed at the definition of a reliable configuration for the automation of the pre-peeling phase, intended as the mutual position of an end-effector holding the coverlay and the grinding tool. The choice is the result of a hierarchical parameter analysis. The first parameter to consider is the Success Index: only configurations leading to 100% successful runs were considered. The Peeled Area PA (and, in particular, the ratio between PA and TA, the Total Area) was assessed as an important discriminating factor for the subsequent peeling completion phase; however, as already stated, all the successful pre-peeling runs were followed by successful peeling completion. From the pre-peeling automation perspective, acquired position standard deviation σ_h is then relevant: since the h value indicates a position suitable for the robot end-effector, a lower σ_h is directly translated in a better repeatability of the process, that represents a fundamental indicator in an automated task. According to this approach, the most suitable set of parameters for the batch of coverlays currently under examination, is represented by the “100” series (a = 10 mm; ϑ = 90°; h = 5.5 mm). It is worth to point out that, according to the discussion reported in Sect. 2.3, in the practical implementation, this particular configuration (ϑ = 90°) does not correspond to the protective film surface tangential to the sanding drum, due to the coverlay intrinsic curvature. As a further remark, in all the successful runs, no damages were observed on coverlay inner surfaces.

In the proposed approach, the process is considered stochastic, relying on the sand particles random distribution. The coverlay/particles contact frequency is directly related to the high rotational speed of the tool. Certain contacts lead to detachment, thus, success probability is strongly increased by high rotational speeds. However, coverlay/tool relative configuration is a fundamental factor to enable the engagement of the only protective film layer; in some cases, indeed, also the coverlay film is engaged itself, leading to a process fail. The experimentation led to the optimal configuration for the specific case-study.