Method for inspecting a correct execution of a processing step of components , in particular a wiring harness , data structure and system

Modern automobiles are now becoming more like computer on wheels . A large amount of signals ( from its own sensors or from the environment like other automobiles or traf fic signs or the like ) are constantly processed and returned to the vehicle in the form of control commands . A prerequisite for this is the correct use of cabling of the individual modules , sensors and actuators , which has become an increasing challenge for automotive production in recent years .

The result is an individually designed wiring harness for more or less each vehicle .

An automotive wiring harness composes of a variety of components , such as standardi zed and special wires , connectors , clips and fixing elements , tubes , tape , cable channels , connector terminals , grommets , etc . The manufacturing of the wiring harness is inflicted by a high degree of complexity . Each wiring harness is customi zed to customer orders and therefore a unique product . This results in a vast number of wiring harness product configurations , which have varying structured bills of material and require di f ferent sequence of production steps and tasks in the manufacturing .

The complexity for wiring harness manufacturing not only stems from the product variety, but also from the highly manual work required to produce a wiring harness . Accordingly, the production of a wiring harness is characteri zed by a high degree of manual labor, especially in the final assembly area of wiring harnesses . In the final assembly operators manually add components and pre-manufactured subassemblies to a wiring harness so called formboard assembly station .

By performing additional manual tasks such as routing wires , insertion of wires into cavities of connectors , taping wires to achieve wire bundles , and so on, the final product , a wiring harness , is produced . Due to up to 90% manual work, failures occur . Human labor results in often random instead of systematic failures and is not as consistent and reproducible as automated processes . This results further in the necessity for a high amount of testing .

Testing uses also the principle of Optical Inspection which provides a powerful , but in itsel f complex means . This is due to complex engineering, which makes it hard to adapt it to changing products , and a high sensitivity with regard to environmental and other influences as light and shadows , variation in component properties , etc . . These are key problems and reasons why this powerful tool is today only rarely used in the field of wire harness manufacturing .

There are already a lot of patent applications published, addressing the wiring harness industry and proposing optical inspection concepts . The following list shows a selection of already known solutions :

In JP2001027518 A, a visual inspection system is proposed based on a so-called pan-tilt camera . That means , the camera is positioned in a predefined location but can capture a wide region of interest by moving along its axis . The camera takes an image of inspection obj ect parts and automatically assesses the quality of the obj ect as good or bad . Whenever a defect is detected, the defect location is indicated with a laser light beam . JP002010249744A describes a system consisting of a moving camera which captures multiple images along the vertical and hori zontal dimension of the formboard . Afterward, the images are stitched together .

From JP002016070710A a moving camera is known, that takes images along a predefined path to determine whether the wiring harness passes or fails visual inspection .

Also in DEI 02016123976B3 a camera is already known from the state of the art , that is mounted on a robot arm . The robot arm moves to multiple monitoring zones to compare the current state of the wiring harness with a target status . I f a deviation is detected, it will be shown on a display and corrected by tools and a manipulator .

There are further patents and applications addressing the inspection of speci fic locations or components of the wiring harness . In JP2007333399A for example , an inspection of the wiring harness components is performed which are held in a holding member . Therefore , the holding member is labeled with an identi fication mark as a reference point for image capturing . The image is processed to assess whether the obj ects seen on the image comply or deviate from predefined requirements .

JP2016223869A, focuses on the detection of tubes by binari zing the image data and a threshold for brightness .

JP002018084542A presents an optical inspection solution to detect holding devices in a predefined region and to determine whether the holding device is part of the wiring harness . A lot of ef fort has been put into image acquisition, recognition of obj ects and the setup of " stages" for the presentation of the obj ect to a camera like gimbals , robots , gantries and the like .

To illustrate the situation more clearly, the following briefly describes a system according to the state-of-the-art , which provides a starting point for the invention described in section three . An exemplary state of the art application machine with a vision system is depicted in Figure 5 .

The optical inspection system consists of a camera 54 for grayscale images , a wiring harness that is mounted on a formboard 53 , and a data processing pipeline to process the image data . The goal of the system is the detection, classi fication, and quality assessment of wiring harness components .

The common use of markers 55 addresses components that look-alike and are partially occluded either by other components due to the geometry of the wiring harness or by the connector holder . For the wiring harness , especially connectors 11 are similar-looking components and often occluded by connector holders . Accordingly, connectors 11 look alike from the outside regarding color and shape but have a di f ferent number of cavities .

Furthermore , they require big connector holders that can hold them on the formboard . For occluded and look-alike components like connectors , feature detection does not perform as reliable because the features are not distinct across components . To solve this problem, more unambiguous and visible features in form of data matrix codes 55 are added . The traditional way to apply machine vision to markers is their attachment to components .

In the state of the art , i f the optical inspection of this scene had to be modelled, models of all the components and the holder had to be generated and trained in advance , the smallest of changes would mean a repetition of a training process starting and based on the training data 51 and the trigger data 52 in the PCS , MES , or the like data storage , where the combination of the information comes to Detection Results and are played back into the data base 52 .

It is the task of the invention to provide a solution to the problems described above .

The main goal of the proposed method and system is to perform an optical inspection of wiring harness states between individual working stations in an automated way .

The problem is solved by a method, executing the steps of claim 1 .

A method for inspecting a correct execution of a processing step of components in wiring harness assembly, by a first inspection program recogni zing a component via an optical sensor by the use of at least one marker on at least one component for simultaneous a) identi fication of the type of the component , and b) identi fication of the location of this component within the wiring harness , c ) the definition of dependencies between identity and location, characteri zed in d) the automated detection and evaluation of the dependencies before the processing step is executed and after the processing step is executed .

The problem is further solved by a data structure representing the workpiece , consisting of the features of claim 9 .

The problem is solved by a system consisting of the features of claim 10 .

Further advantageous embodiments of the invention are stated in the subclaims . It is one basic issue of the invention that a key enabler for Optical Inspection has not received attention yet : the dynamic use of individual marks like fiducials , barcodes , data matrix codes ("markers" ) in the scenes , beyond their basic presence , meaning the change of situation between a first processing status and a subsequent processing status in the processing station .

The invention is defined by the use of the same markers simultaneously for a) the identi fication of components ("comp" ) b) the identi fication of locations (" loc" ) and c ) the definition of dependencies between identity and location (valid, invalid) d) the automated detection and evaluation of the dependencies .

This will be explained in more detail later on the basis of Figure 2 .

The claimed idea is characteri zed by the fact that it uses a dynamic understanding of the implementation of discovery of markers during time , i . e . to make use of the change of visibility of the markers before and after a processing step and the spatial changes of the manufacturing situation .

The invention is represented by the accompanying figures , whereby the figures show as follows :

Figure 1 Elements and states of a wiring connection during manufacturing,

Figure 2 an Example for the use of markers according to the invention,

Figure 3 use of a topology map, and

Figure 4 a map of inspections for a particular point in time , and

Figure 5 a known application of a machine vision system in wire harness manufacturing ( e . g . a final assembly station) . Figure 1 shows a typical sequence in the manufacturing of a wiring connection, used in the processing of a wiring harness, the example not having a restrictive effect on the claimed invention. In a first step 15 a bundle of cables 12 is attached to a connector 11. In the next step, the workpiece 16 is taken to fix the connector 11 in a holder 13, 17. Finally in the last step 18, the bundle of cables 12 is fixed with a tape 14.

In the state of the art, like shown in Figure 5, if the optical inspection of this scene had to be modelled, models of all the components and the holder had to be generated (e. g. for PCS, MES, ...) and the recognition of the components via the optical sensor (camera) 54 will be trained in advance, meaning that the smallest of changes would as a consequence lead to a repetition of the training process.

The use of markers would simplify this process inasmuch, as minor changes to the appearance of the components could be ignored (also possibly caused by different light or shadow) , as long as the marker is visible.

As already mentioned before, Figure 2 provides examples for the use of markers according to the claimed invention.

These are only exemplary scenarios, that means all examples of markers, especially the used terms are only for exemplary means and do not restrict the inventive idea.

"Covering Location with Component" 21

The example shows in the top the view of an empty holder 13, with a space 111 designed to accommodate a connector 11, with a marker Z1 within this empty space, for its location (Marker "Code Loc_Z_l") . Once a connector 11 is inserted in the space 111 of the holder 13, like in the lower left corner, 21, the location code Z1 is no longer visible but in its place the code Al (Marker "Code Comp_A_l") of the component 11. This permits to check, e. g., if all empty holders have been filled in processing step, and if they have been filled correctly, that means with the correct component .

"Wrong Connector in Holder" 22

The next example shows on top an empty holder 13, with a space 112 designed to accommodate a connector 11. Several

( sub- ) locations Z23 (with Marker "Loc_Z_23") , Z45 ("Loc_Z_45") and Z67 ("Loc_Z_67") have been indicated. Once a wrong or misaligned connector 113 is inserted (on the bottom of 22) there are still some of the markings visible, in this example it is Z45 Loc_Z_45. This permits to recognize an error and even to some extent to characterize the nature of the error.

"Identification of wire orientation" 23

The next example shows on top a wire 12 with a single marker Bl ("Comp_B_l") providing of an identification; the use of two markers (bottom) ("Comp_B_23") B23, and ("Comp_B_45") B45 provides additional information about the wires' orientation , if needed (marker "a" above "b") .

"Correct extent of taping" 24

The last example 24 of Figure 2 on the right side shows another use of markers on the wire 12, the possibility to determine if the taping 14 (also: foaming in, shrink wrap, ...) has been applied to the right extent as marker B45 has been made "invisible" by the covering material, and marker B23 is still visible.

With the aforementioned examples it should become clear:

The use of markers to identify empty locations is advantageous in various manufacturing steps (e. g. for storage, transport, fixture for impending operations, assembly, etc.) , thus defining the "key topology" for the manufacturing process. The use of markers is used not only for the identification of components but also for the determination of their orientation.

The use of markers is essential to determine the fulfillment of given process specifications (e. g. extent of tape coverage) in the processing step.

Rules are used that are corresponding with process steps, e. g. o upon arrival at the station, all holders have to be empty (all holder's location labels must be visible) , o upon departure from the station, all holders must be correctly filled (predefined holders' labels must not be visible) ) .

Further, the following elements are proposed, like depicted in Figure 3 and 4 :

A Topology Map 33 indicates all technically meaningful locations for the processing on the formboard.

A Bill of Components 31 lists all relevant parts needed for the manufacturing (of the Wire Harness) .

A List of Rules Rl, R2,..., defines the dependencies between components, locations and manufacturing process effects.

A Map of Inspections (shown in Figure 4) shows on which locations the aforementioned rules Rl, R2, ... have to be applied.

Figure 4 shows the Map of Inspections for a particular point in time in the manufacturing process; e. g. for use after completion of final assembly where every rule Rl, R2, R...is to be applied in the respective location Loc_A_l, Loc_B_l....

In the following, there is shown an exemplary List of Rules Rl, R2, R3, ..., defining exemplary evaluation rules:

The map of inspections from figure 4 can be derived from the topology map, which then is complemented with the rules for the visual inspection that shall be executed at the respective locations . For each assembly process step, predefined rules are defined and will be checked, to monitor the assembly state of the wiring harness .

The following symbolic algorithm describes the execution in both the process engineering and the process execution .

Marker-based optical inspection algorithm

Algorithm for automated optical inspection based on dynamic marker positioning in wiring harness assembly line

During wiring harness development

Initiali ze rule IDs Ry = { la, Cb, Lc, Ed } with y, a, b, c, d eB

Describing the optical inspection issue I

The numerical information of marker on component C, The numerical information on marker at location L, And the effect E as the relation between location and component

Initialize assembly line working station WSx with X G/R

Assign relevant rule IDs Rt £ Ry to WSx = {Rt} with t G/R

During Manufacturing

For WSx

Take image with binary information after assembly task Image processing by camera to identify markers and read numerical information on markers

Checking of actual and target state by MES

If assigned rules WSx = {Rt £ Ry} comply with numerical information provided by camera

Then quality OK and WS x+1

Else quality NOK, visualization of failure and rework

Endif

Endf or

In the following, a further application example for the application and detection of codes in wire harness manufacturing is provided.

The system bases on two different inspection programs. For fully visible components on the image, the inspection program relies on component modeling whereas data matrix codes are used for partially visible and similar components.

The first inspection program addresses wiring harness components that possess distinct features and therefore, can be differentiated visibly. Examples for such components are different types of clips, foamed parts, and relay boxes.

For the detection of these components, virtual component models are created. The models contain component-specific information such as edges and component color . Creating models and choosing relevant features for the models , also referred to as feature engineering, is conducted for each component , which should be optically inspected in wiring harness manufacturing .

In the manufacturing, the models are the foundation for obj ect detection and quality assessment . The camera captures an image and the image data processing pipeline starts .

For each model , the image is scanned for obj ects that are similar to the models . However, the obj ects do not have to be of the same si ze and orientation because the data processing algorithm can consider scaling and full rotation of the obj ect .

I f a critical threshold is passed, which means i f enough found features align with the model , the found obj ect is marked with a bounding box on the image . The rotation of the found obj ect in relation to the model is given in degrees and the location of the bounding box in relation to the bottom left corner as the coordinate origin is given in x and y coordinates . There is a final visuali zation of what kind of obj ects were found, the location and rotation of each obj ect , and the number of each component in the image .

In one advantageous embodiment of the invention, the data matrix code applied are according to ECC 200 ( ISO/IEC 16022 : 2006 ) . In the example shown, the data matrix code is positioned into the connector holders as well as on the connectors . Due to lack of space in the component holders , the data matrix codes have a si ze of 10 columns and 10 rows . Accordingly, the codes can contain numerical information from 0 to 255 . The advantage of data matrix code is that the information on them is condensed and they can be read even i f they are partially damaged . Instead of detecting the obj ect straight , the camera detects the data matrix code and provides the information that the code contains .

The information of the data matrix code positioned in the connector holder is used to derive the existence of a connector in a connector holder .

The information of the data matrix code on the connector is used to identi fy and classi fy the mounted connector and to assess its position on the formboard .

In the industrial implementation, the camera captures an image and scans the image for data matrix codes . When data matrix codes are detected, the final visuali zation shows codes found in form of numbers .

To enable both inspection programs , especially the second approach, data speci fication and data preprocessing during the development process is important .

During wiring harness design, once components are chosen, models of the obj ects must be established for the first inspection program . I f the second approach is chosen, data matrix codes must be speci fied by designers during formboard design . Accordingly, data matrix code codes are generated and placed into connector holders in the formboard design .

It is important that each formboard contains unequivocal data matrix codes and is not labeled with the data matrix codes with the same information twice .

The information the data matrix code contains is paired with the component , that should be placed into the connector holder, e . g . connector type 123 is associated with the data matrix code 006 in the connector holder that is positioned at the location ( 245 | 329 ) or, like shown in Figure 3 , Loc_A_l , Loc_B_l , .... Then during line balancing, the working station for optical inspection is speci fied and the testing steps are integrated into the working instructions as well . The mapping of data matrix code and components or models and components , more precisely the information which components and quality characteristics were supposed to be identi fied in which working station, is given to the manufacturing execution system (MES ) .

During the manufacturing, the camera captures images according to the inspection programs :

For the first inspection program the camera outputs which and how many components were found and their quality characteristics , precisely component name , location, and rotation . The output of the camera is then matched by the MES system with data generated during wiring harness development . I f deviations are noticed, then failures can be deducted . Failures can be e . g . missing component , wrong component in a speci fic location, and component mounted in wrong geometry .

For the second inspection program the camera outputs numbers that were detected based on data matrix codes and this information is given to the MES system . The Manufacturing Execution System (MES ) system processes this information and derives the quality characteristics . All connectors are mounted correctly i f data matrix codes of correct connectors have been identi fied, data matrix code that are concealed by connectors have not been seen, and that empty connector holder and their data matrix codes are also read by the camera .

I f a wrong data matrix codes is outputted by the camera, the MES will conclude that connectors are missing, have been mounted in the wrong connector holder, or wrong connectors have been mounted .

The optical inspection is conducted on or after each manual working station . The system, which is the camera in conj unction with a Manufacturing Execution System (MES ) that has been provided with the right information and is able to dynamically inspect a high variety of wiring harness configurations .

In case the assembly station or formboard is too big for one image e . g . due to the technical properties of the camera or lack of space resulting in a short distance between formboard and camera, in another advantageous embodiment of the invention, multiple regions of interest for optical inspection can be defined .

Accordingly, the formboard is structured into a reasonable number of regions of interest which are speci fied during the step of designing the map of inspection . The inspection based on dynamic marker change is then conducted for each of the region of interest allowing the quality assessment of big as well as small wiring harness .

The main goal of the proposed method and system is to perform the optical inspection of wiring harness states between individual working stations in the final assembly in an automated way . Proposed solutions with moving cameras cannot be applied here because moving cameras or cameras mounted on robot arms require safety zones in order not to harm the workers in this area .

By using simple industrial cameras , the costs associated with this solution are reduced because no additional hardware is required to move the camera .

Moreover, the solution proposed addresses the optical inspection of a high component variety . The creation of component models with feature engineering and data matrix codes in conj unction with the data flow from wiring harness design over the Manufacturing Execution System MES system to the camera in the field enables the detection of di f ferent components in the manufacturing on the formboard dynamically . In contrast , the state of the art , discussed earlier in this document is purely focused on the detection of single components at one moment in time . The ef fort associated with model and data matrix code creation is comparably low because the models and codes can easily be reused .

Furthermore , the dynamic change of markers allows quick assessment of a component ' s presence and location in the respective assembly process step . In case of assembly failures , which means that predefined inspection rules at assembly process steps are not ful filled, the failure becomes evident and can be resolved immediately . Currently, the quality of wiring harnesses is checked at the end of the line and rework is highly time-consuming and laborious .

The proposed automated optical inspection system can be implemented for all rigid and deformable components in the wiring harness industry . Especially in the light of unfolding trends such as autonomous driving and electri fication, the monitoring and quality evaluation of safety-critical components is important . As a result , either all components during design could be chosen for modeling and labeling with data matrix codes . Alternatively, chosen components can be included for optical inspection, e . g . safety-critical components with a high AS IL level , quality critical components that required high rework and quality issues in the past should be modeled and labeled .