METHOD AND CONTROL UNIT FOR CONTROLLING THE PRODUCTION OF A PRODUCT

Description

[0001] The present invention relates to a method for controlling the production of a product according to claim 1 and a control unit for controlling the production of a product according to claim 10.

[0002] Distributed production processes are typically carried out on the basis of statistical quality management, using, for example, the Six Sigma method for process improvement. This places very high demands on the precision of the individual components, meaning that the corresponding production processes are also subject to high precision requirements. In the Six Sigma method, the width of the tolerance range for a characteristic of a component corresponds, for example, to a spread with a standard deviation of approximately six. Based on such a spread, one also speaks of zero-defect production. If the individual complementary components are combined randomly or without selection during the manufacture of products, considerable quality requirements exist in the supply chain.This is primarily due to the fact that product quality deteriorates under statistical process control, such as the Six Sigma method, according to Gaussian error propagation. Error propagation places high demands on low dispersion of the characteristics of the complementary components, thus low variation. The manufacturing tolerances in the production processes of the complementary components must be correspondingly low.

[0003] By specifically combining complementary components, however, it is possible to achieve greater variation in the characteristics of the complementary components, or to increase manufacturing tolerances during production. This requires determining the characteristics of the independently produced complementary components through measurement processes for each component and assigning them to the respective components. For example, the length of a component is measured and stored in a storage medium in association with an identifier for that component. Based on the measured characteristics, the complementary components can be specifically combined in a combination step, so that a selected combination of components forms a product that meets the required quality.Instead of producing all complementary components of a product as precisely as possible, the targeted selection of combinations allows for increased variance in the production of the components. This allows individual components to be produced with lower quality but significantly cheaper. A method for the targeted combination of complementary components in automated product production is described, for example, in patent application DE 102016 121 092 A1. "Quantum annealing - Wikipedia", December 16, 2019, XP055700706, describes the process of the same name.

[0004] The optimal combination of complementary components requires the evaluation of all possible combinations of complementary components. The computational effort required for this increases significantly with the number of complementary components to be combined. For example, if two sets of complementary components are combined, each with N components, the number of possible combinations is the factorial of N. For example, with N = 90 components, 90! = 10^138 combinations are possible, or rather, must be evaluated. Determining an optimal combination of several complementary components accordingly requires a considerable computational effort, which requires corresponding resources (computing infrastructure, time, etc.).

[0005] In particular, the combination problem for large numbers of components to be combined is not solvable for classical digital computing units.

[0006] The object of the underlying invention is to provide a method that enables the efficient, simple, and cost-effective manufacture of products that have complementary components. Furthermore, the object of the invention is to provide a control unit that can at least partially execute the method.

[0007] The underlying problem is solved by a method having the features of claim 1. Advantageous embodiments of the invention are given by the subclaims.

[0008] The method according to the invention provides that the combination step is carried out at least partially by means of a quantum-based computing unit, wherein the selected combinations of the components, as a whole, correspond as closely as possible to a target value. The solution to the combination problem, or the finding of an optimal combination, is carried out according to the invention using the quantum-based computing unit. The selected combinations forming the products, or optimal combinations, are selected by means of the quantum-based computing unit such that the features of the components complement each other such that the entirety of the products correspond as closely as possible to a target value, or have a specific quality.

[0009] The use of a quantum-based computing unit has the advantage that the optimal determination of selected combinations of complementary components is possible in a shorter time. In particular, using quantum-based computing units, optimal combinations can also be determined for a larger number of complementary components.

[0010] According to the invention, the quantum-based computing unit is a quantum annealer or a digital annealer unit simulating a quantum annealer. Quantum annealers and digital annealer units simulating a quantum annealer, or quantum-inspired digital computers, manage to provide a solution to optimization problems by exploiting quantum mechanical effects, such as the tunneling effect or emulations thereof. Advantageously, this also achieves an energy-efficient solution to the combination problem, since quantum annealers, in particular, have minimal energy consumption for computation, except for the cooling power.

[0011] Quantum annealers are optimization machines that use so-called quantum bits, or qubits, for programming. At the beginning of the quantum annealing process, all qubits are brought into a quantum mechanical state of superposition. The magnetic field required for this is continuously reduced during the optimization process, while simultaneously the qubits are manipulated to implement the combination problem. Implementation here means controlling the qubits via additional external magnetic fields and adjusting the coupling strength between the different qubits. During the subsequent annealing optimization process, the qubits align themselves in such a way that the total energy of the physical system is minimized. At the end of the process, the qubits are in the classical state and can thus be unambiguously read out. The result clearly represents the connections between the selected combinations of components.Quantum annealers, for example, are manufactured by the company DWave.

[0012] Preferably, the quantum-based computing unit is spatially separated from a control unit intended to control the production process and communicatively connected to it. This applies in particular to the quantum annealer, where the computing power of the quantum annealer is provided by a cloud service. This has the advantage that the quantum-based computing unit, especially the cost- and maintenance-intensive quantum annealer, is outsourced from the production facility.

[0013] Digital annealer units simulating a quantum annealer are currently serving as a bridging technology. These digital annealer units perform classical thermal annealing but are capable of emulating the quantum mechanical tunneling effect by injecting additional temperature during the saturation of the classical algorithm. To achieve this, the system is allowed to sporadically make larger changes towards the end of the classical annealing process than would be possible with a purely classical method. While a quantum annealer can reach a minimum behind a potential hill, where there is a probability of "tunneling through" the hill, the digital annealer achieves this goal via the briefly higher temperature, thus classically "moving over the potential hill." In theory, both technologies achieve a deeper minimum.For example, a digital annealer unit from Fujitsu is currently capable of solving a combination problem with a problem size of N=90, making it suitable for use in the manufacturing industry. Digital annealer units have the advantage of being cost-effective and spatially compact, making them ideal for integration into a production facility or for use in a control unit within the production facility.

[0014] According to the invention, a function of the characteristics of all possible combinations of the components is determined, whereby the function describes the complementary relationship between the characteristics. For a product in which the components A_i with the characteristics a_i are combined with the complementary components B_j with the characteristics b_j, the function can be written as f(a_i, b_j). The function f(a_i, b_j) describes the complementary relationship between the characteristics a_i and b_j, where the characteristics are, for example, dimensions, weights or material properties. In a simple form, the function is formed, for example, by adding two lengths and has the form f(a_i, b_j) = a_i + b_j. Advantageously, the use of the function can be used to describe the relationship between the characteristics, so that the effects of varying the characteristics can be determined easily.

[0015] According to the invention, variances of all possible combinations of the components are determined in the combination step, with the respective variance corresponding to the difference between the respective result of the function and the target value. The target value is also referred to as the Key Performance Indicator (KPI). For a combination of components A_i and B_j with characteristics a_i and b_j, the corresponding variance would be the result of the function f(a_i, b_j) minus the target value KPI. If characteristics a_i and b_j of the components do not deviate from the target value, the result of the function corresponds to the target value. In ideal production of the components, the variance with respect to the target value is therefore zero. This is theoretically possible but very unlikely if the characteristics vary due to production.By determining the variances, one advantageously obtains a statement about the extent to which the respective combination of components corresponds to the target value.

[0016] A preferred embodiment of the invention is characterized in that the determination of the respective results of the functions and / or the variances is carried out using a conventional digital computer unit. The computing power of a conventional digital computer unit is typically sufficient for calculating the variances of all possible combinations of complementary components. The use of a conventional digital computer unit has the advantage that it is cost-effective and can be easily implemented within a production plant, preferably as part of a control unit.

[0017] According to the invention, the variances of all possible combinations of the components are each assigned a quantum bit as a binary quantity. The quantum bit, or qubit, can have the values 0 or 1 in the classical state and is referred to below as x_ij. This advantageously allows the implementation of the underlying combination problem on a quantum-based computing unit.

[0018] A preferred embodiment of the invention provides that an objective function is used in the combination step that corresponds to the minimum of the square sum of the quantum bits of all possible combinations. Such an objective function advantageously enables implementation on a quantum-based computing unit.

[0019] A preferred embodiment of the invention provides that constraints are used in the combination step that take into account that all components are combined with one another. These constraints ensure that each component is combined with a complementary component.

[0020] In particular, for a product that has two complementary components, a first constraint ensures that each component A_i is assigned a complementary component B_j. A second constraint ensures that each component B_j is assigned a complementary component A_i. The use of both constraints advantageously ensures that a unique assignment between the components is achieved.

[0021] A preferred embodiment of the invention provides that the constraints are weighted during the combination step using a scaling factor. In the following, the scaling factor, or Lagrange parameter, is referred to as λ. Using the scaling factor, the influence of the constraints can advantageously be variably adjusted.

[0022] A preferred embodiment of the invention provides that a Quadratic Unconstrained Binary Optimization (QUBO) is used in the combination step, which is a sum of the objective function and the constraints. During the combination step, the Quadratic Unconstrained Binary Optimization (QUBO) is performed using the quantum-based computing unit, whereby the quantum-based computing unit determines the minimum of the sum and outputs as a result the selected combinations of components that, as a whole, best correspond to the objective value. The Quadratic Unconstrained Binary Optimization (QUBO) comprises quadratic combinations of qubits, so that the determination of the optimal combinations can advantageously be carried out efficiently using a quantum-based computing unit.

[0023] According to the invention, the components are transported to a joining unit in a transport step in order to be assembled into the respective product in a joining step. Identifiers are stored in a data set readable by the joining unit using identification means of the components of the selected combinations. The data set is stored, for example, in a storage means of the control unit, in particular wherein identifiers of the components of a selected combination are stored in association with one another. For example, the identifiers are serial numbers or other alphanumeric identifiers, and the identification means are barcodes, QR codes, or RFID tags. Storing a data set containing the identifiers of the selected combination enables the joining unit to combine a component with the corresponding complementary component selected in the combination step to form a product.

[0024] A preferred embodiment of the invention provides for the selected combinations of components to be transported together as pairs. For example, the selected combinations are transported within a transport means to the joining unit and assembled there to form a product. This has the advantage that the optimal combinations of complementary components are spatially separated from one another, preventing an unintentional combination of complementary components.

[0025] A preferred embodiment of the invention provides that the components are provided with an identifier that can be identified by the joining unit and indicates their complementary components of the selected combinations. The identifier is preferably an RFID tag in which the identifiers of the complementary components of the selected combination are stored. The identifier can be read by the joining unit, so that the joining unit knows which component is to be joined with the respective complementary component. Advantageously, the components can be transported independently of one another to the joining unit and joined there to form the desired product, in particular without the joining unit having to access the data set containing the identifiers of the components of the selected combinations.

[0026] The object of the invention is also achieved by a control unit for controlling the production of a product, which is designed to carry out the method described above.

[0027] In the following, preferred embodiments of the present invention are explained with reference to the figures: Fig. 1: shows a schematic representation of the combination of two complementary components A_i and B_j of the sets A and B, Fig. 2: shows a schematic representation of the method according to the invention for controlling the production of a product, and Fig. 3: shows a schematic representation of a production plant that carries out the method according to the invention.

[0028] In the following, features of the present invention are explained using preferred embodiments.

[0029] Fig. 1shows, in the form of a mathematical graph, an example of the combination of two complementary components. The first set A has a number N of first components A_i. These are combined with the second components B_j of the second set B to form a product. The dashed lines indicate all possible combinations, while the solid lines represent the selected combinations, or the optimal combinations. The first component A_i has a feature a_i that is complementary to the feature b_j of the second component B_j and exhibits a production-related variation. The feature b_j also exhibits a production-related variation.The aim of the invention is to provide a method that allows an optimal combination of components A_i and B_j based on the characteristics a_i and b_j, so that the product formed from components A_i and B_j meets a specific quality requirement. For example, the components are components that are combined as a product to form a component group, or pure substances that are combined as a product to form a chemical compound. The characteristics can be, for example, dimensions, weights, or material properties.

[0030] Fig. 2shows a schematic representation of the method according to the invention, wherein the first components A_i are manufactured by means of a first production process 1a and the second components B_j are manufactured by means of a second production process. Due to production reasons, the features a_i and b_j of the components A_i and B_j exhibit variations. The features a_i of the components A_i are determined by means of a first measuring process 2a and the features b_j of the components B_j are determined by means of a second measuring process 2b. For example, a dimension, a weight or a material property is determined by means of the measuring processes 2a and 2b. The components A_i and B_j have an identification means 13 which has a unique identifier of the respective component A_i and B_j. The identification means 13 are, for example, barcodes, QR codes or RFID tags, wherein the identifier is, for example, a serial number or another alphanumeric identifier.The characteristics a_i and b_j of components A_i and B_j determined by the measurement processes 2a and 2b are stored in a storage medium in association with the respective identifier. In particular, the storage medium is a storage medium 16 of a control unit 10. Alternatively, the respective characteristic a_i can also be stored in a storage medium arranged on the component. The complementary components A_i and B_j are combined with one another in a combination step 3 based on the measured characteristics a_i and b_j, so that a selected combination of components A_i and B_j each forms a product.

[0031] The combination step 3 is at least partially carried out by a quantum-based computing unit 18, wherein the selected combinations of components A_i and B_j, as a whole, correspond as best as possible to a target value, or a Key Performance Indicator (KPI). As an example, the following describes the determination of a Quadratic Unconstrained Binary Optimization (QUBO), which is used for programming a quantum-based computing unit 18, for the problem from Fig. 1 shown.

[0032] For each component A_i with the respective characteristic a_i, exactly one complementary component B_j with the respective characteristic b_j is searched, so that the expression ∑ j x ij f a i b j − KPI 2 is minimized. The combination of an a_i and a b_j is referred to as x_ij, where x_ij is a binary number with the value 0 or 1 and is implemented using a qubit when using a quantum-based computing unit 18. The function f(a_i, b_j) is any function that describes, or relates, the complementary relationship between the characteristics a_i and b_j of the components A_i and B_j. The above expression calculates, for each a_i, the variance of all combinations with b_j, 1 < j < N, and the variable x_ij (0 or 1) decides whether this combination contributes to the sum. If the above expression is summed over all elements of the set A and the result is minimized, the following objective function for the combination problem is obtained. min ∑ i ∑ j x ij f a i b j − KPI 2 .

[0033] For all combinations of components A_i and B_j, the respective results of the function f(a_i, b_j) are determined depending on the respective characteristics a_i and b_j. This is preferably done using a conventional digital computer unit 17, which is in particular a component of the control unit 10. The variance, i.e., the difference between the respective result of the function f(a_i, b_j) and the target value KPI, is preferably also determined by a conventional digital computer unit 17.

[0034] By introducing constraints, all components A_i and B_j are combined. In particular, the minimization of the above expression ensures that not all x_ij vanish simultaneously. This requires that exactly one b_j be assigned to all a_i. The following then applies: ∑ j x ij = 1 , which is in the following expression ∑ i ∑ j x ij − 1 2 = 0 for the first constraint. This quantity can be understood as a penalty term for non-compliance with the constraint. If an a_i is connected to none or to several b_j, the constraint provides a contribution - only in the case of a unique combination (A_i, B_j), or a unique pair (A_i, B_j), does the contribution of this term disappear. During the optimization process, the algorithm will therefore specifically assign each component A_i of the set A to exactly one complementary component B_j of the set B.

[0035] The quadratic expression is simplified and yields ∑ i ∑ j x ij 2 + 2 ∑ j ∑ k > j x ij x ik − 2 ∑ j x ij + 1 = 0 .

[0036] With x ij 2 = x ij for binary variables one obtains ∑ i ∑ j x ij + 2 ∑ j ∑ k > j x ij x ik − 2 ∑ j x ij + 1 = 0 and ∑ i 2 ∑ j ∑ k > j x ij x ik − ∑ j x ij + 1 = 0

[0037] Due to the symmetry of the task, it must be ensured that exactly one a_i is assigned to all b_j. Thus, the following applies: ∑ i x ij = 1

[0038] The latter expression is transformed to ∑ j ∑ i x ij − 1 2 = 0 .

[0039] Following the derivation of the first constraint, one obtains the following expression for the second constraint: ∑ j 2 ∑ i ∑ k > i x ij x kj − ∑ i x ij + 1 = 0

[0040] The expression for the Quadratic Unconstrained Binary Optimization (QUBO) of the combination problem consists of the sum of the objective function and the two constraints: min ∑ i ∑ j x ij f a i b j − KPI 2 + λ ∑ i 2 ∑ j ∑ k > j x ij x ik − ∑ j x ij + 1 + λ ∑ j 2 ∑ i ∑ k > i x ij x kj − ∑ i x ij + 1 where the constraints are weighted by a scaling factor λ, the so-called Lagrange parameter. The larger the Lagrange parameter, the greater the influence of the constraints during the optimization. The QUBO can ultimately be simplified to expression (12): min ∑ i ∑ j x ij f a i b j − KPI 2 + λ ∑ i ∑ j ∑ k > j x ij x ik + ∑ j ∑ i ∑ k > i x ij x kj − ∑ i ∑ j x ij where the constants that have no influence on the minimization have been omitted.

[0041] Quadratic Unconstrained Binary Optimization (QUBO) is performed using a quantum-based computing unit 18. As a result of the computation process, the qubits are in the classical state and can thus be unambiguously read out, yielding the selected combinations of the components A_i and B_j. The solution determined by the quantum-based computing unit 18 does not necessarily correspond to the absolute minimum of the combination problem. However, in the presented method, the "second-best solution" is also sufficient, or rather, this is accepted due to the performance of the quantum-based computing unit 18.

[0042] Components A_i and B_j are transported to a joining unit 15 in a transport step 4 to be assembled into the respective product in a joining step 5. In order for the selected combinations of components A_i and B_j to be used to assemble the respective product in joining step 5, the selected combinations must be known to the joining unit 15. For this purpose, the identifiers of the identification means 13 of components A_i and B_j of the selected combinations are preferably stored in a data set accessible to the joining unit 15. For example, the data set can be in the form of a table, with the identifiers of components A_i and B_j of the respective selected combination being assigned to a product.

[0043] Preferably, the selected combinations of the components A_i and B_j are transported together as pairs in a transport unit to the joining unit 15 in transport step 4.

[0044] Preferably, components A_i and B_j are provided with an identifier that can be identified by the joining unit 15. For example, the identifier of component A_i is provided with the identifier that specifies the complementary component B_j of the selected combinations. In this way, the joining unit 15 can determine, by reading the identifier, which components A_i and B_j are to be combined to form a product. The identifier is preferably an RFID tag.

[0045] Fig. 3shows a schematic representation of a production plant for manufacturing a product from the complementary components A_i and B_j, wherein the control unit 10 is used to control production. The production plant comprises a first production unit 11a in which the first components A_i are produced. The second components B_j are produced in the second production unit 11b. Both components A_i and B_j are provided with an identification means 13, which has a unique identifier for the respective component A_i and B_j. The features a_i of the first components A_i are determined by means of a first measuring unit 12a. A second measuring unit 12b determines the features b_j of the second component B_j. The respective features a_i and b_j of the components A_i and B_j are stored in association with their indicators in a storage means 16 of the control unit 10. The dashed lines in Fig. 3A communicative connection between the units of the production plant is shown, via which the control unit 10 can control the units and from which it obtains information. For example, the control unit receives the respective characteristics a_i of the components A_i from the first measuring unit. The communication between the control unit 10 and the quantum-based computing unit 18 is also shown in dashed lines. The solid lines between the units represent transport paths of the components A_i and B_j.

[0046] Based on the measured characteristics a_i and b_j, the result of the function f(a_i, b_j) is determined for all combinations using the classical digital computer unit 17. The computer unit 17 is preferably a component of the control unit 10. The classical digital computer unit 17 also calculates the respective variance for all combinations, which corresponds to the difference between the respective result of the function f(a_i, b_j) and the target value KPI. The variances of all combinations determined by the classical computer unit 17 are transmitted to a quantum-based computer unit 18. The quantum-based computer unit 18 performs the previously described Quadratic Unconstrained Binary Optimization (QUBO). The quantum-based computer unit 18 is preferably a quantum annealer or a digital annealer unit simulating a quantum annealer.In the case of a quantum annealer, the latter is preferably spatially separated from the control unit 10, with the computing services of the quantum annealer being provided in the form of a cloud service. In the case of an annealer unit simulating a quantum annealer, the quantum-based computing unit 18 can be spatially integrated into the production facility.

[0047] The components A_i and B_j are transported to a joining unit 15 by means of a transport unit 14, wherein the control unit 10 controls the transport unit 14 and joining unit 15 such that the selected combinations of the components A_i and B_j are assembled to form the respective products. List of reference symbols:

[0048] Afirst set Bsecond set A_first component(s) B_jsecond component(s) a_iFeature of the first component(s) b_jFeature of the second component(s) fFunction, of the features KPITarget value x_ijQuantum bit, Qbit 1aFirst production process 1bSecond production process 2aFirst measuring process 2bSecond measuring process 3Combination step 4Transport step 5Joining step 10Control unit 11aFirst production unit 11bSecond production unit 12aFirst measuring unit 12bSecond measuring unit 13Identification means 14Transport unit 15Joining unit 16Storage means 17Classical digital computing unit 18Quantum-based computing unit

Claims

1. Method for controlling the manufacture of a product that includes at least two components (A_i, B_j) that are complementary to each other, of which one set (A, B) of each has been manufactured in independent production processes (1a, 1b) and each includes at least one feature (a_i, b_j) with a variation induced by the manufacturing process, wherein the features (a_i, b_j) of the components (A_i, B_j) have been determined by measurement processes (2a, 2b) and have been stored in a storage means (16) with assignment to an identifier of the respective component (A_i, B_j), and wherein the complementary components (A_i, B_j) are combined from the sets (A, B) in a combination step (3) based on the respective features (a_i, b_j) in such a way that each selected combination of the components (A_i, B_j) forms a product in each case, characterized in that the combination step (3) is performed at least in part by means of a quantum-based computing unit (18), wherein the selected combinations of the components (A_i, B_j) in their entirety correspond optimally to a target value, in that the quantum-based computing unit (18) is a quantum annealer or an annealer unit that simulates a quantum annealer, in particular wherein the quantum-based computing unit (18) is separated spatially from a control unit (10) intended for controlling production and can communicate therewith, in that in the combination step (3) a function (f(a_i, b_j)) of the features (a_i, b_j) of the components (A_i, B_j) is used that describes the complementary relationship between the features (a_i, b_j), in that in the combination step (3) variants of all possible combinations of the components (A_i, B_j) are determined, wherein the respective variant corresponds to a difference from the respective result of the function (f(a_i, b_j)) and the target value (KPI), in the combination step (3), in that in the combination step (3) a quantum bit (x_ij) is assigned as binary variable to the variants of all possible combinations of the components (A_i, B_j), in that in a transport step (4) the components (A_i, B_j) are transported to an adding unit (15) and in an adding step (5) added to the respective product, wherein the identifiers of the selected combinations of components (A_i, B_j) are stored in a dataset that can be read out by the adding unit (15), and the adding unit (15) adds a component with the corresponding complementary component selected in the combination step to a product on the basis of the identifiers.

2. Manufacturing control method according to Claim 1, characterized in that the determination of the results of the functions (f(a_i, b_j)) and / or of the variants performed by means of a classic digital computing unit (17).

3. Manufacturing control method according to Claim 1, characterized in that in the combination step (3) a target function is used that corresponds to the minimum of the sum of squares of the quantum bits (x_ij) of all possible combinations, in particular wherein the target function is min ∑ i ∑ j x ij f a i b j − KPI 2 .

4. Manufacturing control method according to any one of the preceding claims, characterized in that in the combination step (3) secondary conditions are used which take into account that all components (A_i, B_j) are combined with each other, in particular wherein a first secondary condition is considered, according to which one complementary component (B_j) is assigned to each component (A_i), wherein the first secondary condition with use of a quantum bit (x_ij) is preferably ∑ j ˙ x ij = 1 , and in particular wherein a second secondary condition considers that a complementary component (A_i) is assigned to each component (B_j), wherein the second secondary condition with use of a quantum bit (x_ij) is preferably ∑ i ˙ x ij = 1 .

5. Manufacturing control method according to Claim 4, characterized in that in the combination step (3) the secondary conditions are weighted by means of a scaling factor (λ).

6. Manufacturing control method according to Claims 4 or 5, characterized in that in the combination step (3) a sum of the target function and the secondary conditions is used as Quadratic Unconstrained Binary Optimisation (QUBO), in particular wherein the Quadratic Unconstrained Binary Optimization (QUBO) contains the expression min ∑ i ∑ j x ij f a i b j − KPI 2 + λ ∑ i ∑ j ∑ k > j x ij x ik + ∑ j ∑ i ∑ k > i x ij x kj − ∑ i ∑ j x ij 7. Manufacturing control method according to Claim 6, characterized in that the Quadratic Unconstrained Binary Optimisation (QUBO) is performed by means of the quantum-based computing unit (18).

8. Manufacturing control method according to Claim 1, characterized in that the selected combinations of the components (A_i, B_j) are transported together as pairs.

9. Manufacturing control method according to Claim 1, characterized in that the components (A_i, B_j) are furnished with a code that can be identified by the adding unit (15), which code specifies the identifiers of the complementary components (A_i, B_j) of the selected combinations, in particular wherein the code is an RFID tag.

10. Control unit for controlling the manufacture of a product, wherein the control unit (10) is configured to execute the method according to Claims 1 to 9.