Automatic Configuration of an Order as an Integral Part of a Cyber-Physical System in a Manufacturing Operating According to Mass-Customisation Strategy

Abstract

The contemporary consumer market determines the use of mass customisation by manufacturers. Therefore, from the point of view of Industry 4.0 (I4.0), concept data and communications and analytics are relevant to the implementation of the mass-customisation strategy. The problem involves at least three subjects: how to connect the currently available information system within an enterprise with new I4.0 technologies, how new information solutions can support the verification of the feasibility of tailoring an order to the needs of the customer, and, finally, how to model a knowledge base for a cyber-physical system containing a formal record of the knowledge acquired regarding customer preferences. Therefore, in this paper, we developed a new algorithm that will enable, on one hand, the collection and recording of customer preferences, and, on the other hand, the integration of those data and information that are available within the Enterprise Resource Planning (ERP) system. The main contribution of this work is the use of specialist knowledge and data from ERP regarding production capabilities implemented in a manufacturing enterprise in order to model the scenario of generating possible orders for a client, and finally applying the new solution for the operation of manufacturing according to mass-customisation strategy in a real company that manufactures mattresses.

1. Introduction

Mass customisation is an economic rebalancing strategy dictated by the current competitiveness of the market. It involves manufacturing different product models while reducing costs in terms of tools and equipment. Manufacturers, in line with this strategy, are faced with the challenge of minimising changes to the production processes, quality, machine facilities, number of employees, or time of delivery by simultaneously improving production flexibility and workface flexibility [1,2]. The capability for mass customisation has been identified as an evaluation factor of the producer’s ability to compete in a rapidly changing environment. This type of production requires manufacturers to adapt to the preferences of customers and build close relationships with them in order to gain knowledge of their preferences and requirements [3]. This strategy also takes into account that there is a link between satisfaction and loyalty in the context of the consumer’s co-design experience [4].

The contemporary consumer market determines the use of mass customisation by manufacturers. Mass customisation is an economic rebalancing strategy that consists of the mass production of customised products at standard costs. It involves manufacturing different product models while reducing costs in terms of tools and equipment [1]. Manufacturers conforming to this strategy are faced with the challenge of minimising changes to production processes, to quality, to machine facilities, to the number of employees, and to delivery times, along with the simultaneous improvement of both production and workface flexibilities [1,2]. From the point of view of the Industry 4.0 concept, a dedicated architecture for the cyber-physical system is required to transform the acquired data into useful information [5]. Moreover, the following Industry 4.0 technologies are relevant to the implementation of the mass-customisation strategy: data and communications, analytics and artificial intelligence, human/machine interactions, and automated machinery [6]. As part of data and connectivity, technologies such as the Internet of Things and Cloud technologies, as an example of Industry 4.0 technologies, are used. They are responsible in the supply chain for centralising storage data, processing it, communication, and the transfer of information. These technologies support the customisation of products by centralising management. They facilitate the availability and feasibility of checking customers’ orders and provide, in real time, access to data from all links in the supply chain. They are also used in the production process, where they allow machines to communicate with each other in order to deliver personalised products as quickly as possible [7]. On the other hand, artificial intelligence, in product customisation, can be used to forecast and plan the demand for products by collecting data directly from products sold [8]. The third group of technologies—human/machine interactions in mass customisation, among other things—facilitates the selection process in storage. Therefore, they provide an opportunity to move the production process into distribution, thus facilitating the faster delivery of products to market and increasing the possibility of customising the products supplied [6]. As part of I4.0 technologies, automated machinery, such as 3D printers, is used in mass customisation. Using these 3D printers in production allows products to be produced that are precisely tailored to the customer’s preferences. The only limitations in such cases are the types of raw materials from which items can be printed [8]. According to the mass-customisation production strategy, mixed-production-line models are used, which allow different models of the basic product’s variant to be produced without high costs. In addition to the diversity of products, such production lines allow a quick response to changing customer expectations and requirements [4,9]. This is, however, associated with certain limitations. A large number of raw materials must be ready to hand at the production line, as any shortage of parts will cause both the line and production to stop [10]. However, regardless of the many possibilities of the usage of I4.0 technologies within production, the problem is how to use them in order to simplify the customisation process. Therefore, based on an analysis of the literature on the subject, we determined that there is a need to develop a solution that will enable the collection and recording of customer preferences and data and information that are available within the manufacturing information system, namely the Enterprise Resource Planning (ERP) system, in order to support the completion of customised orders from clients. Therefore, there is a research niche in the area of acquiring and storing specialised data and information about customer expectations, converting these expectations into technical parameters of products as an integral part of the Cyber-Physical System (CPS) currently applied or planned to be applied in the company to dynamically and quickly meet the requirements [11].

Therefore, the goal of this research is to develop a new algorithm as an integral part of a CPS that will enable the adjustment of previously manufactured products or new configurations of products to meet the customer’s expectations concerning the knowledge about production capabilities that is available within an ERP system.

A CPS can be defined as an integration of computational and physical processes, where computers and networks monitor and control physical processes, usually through feedback loops, where physical processes influence computation and vice versa [12]. In the CPS, the physical part provides access to data, along with sensors and communication systems to collect information about the real world and communicate with the cyber layer, which further analyses and sends the results to the appropriate physical systems via many feedback loops [13].

Therefore, the main contribution of our work is visible in three aspects: (1) modelling a knowledge base for a Cyber-Physical System containing a formal record of the knowledge acquired regarding customer preferences owing to the use of rules for converting expert knowledge and data from an ERP system regarding production capabilities implemented in a manufacturing enterprise; (2) the development of an algorithm for automatic customised order configuration; and (3) the application of our solution as a part of a cyber-physical system of a real company that manufactures mattresses according to the mass-customisation strategy.

2. Materials and Methods

2.1. Design Methodology

The role played by the CPS applied within a manufacturing process operating according to the mass-customisation strategy and the need for automatic customised order configuration point to the objective of this paper, which is to explore the following research questions (R):

R1. How should some knowledge about orders, products, materials, product characteristics, and customer expectations within manufacturing processes operating according to the mass-customisation strategy be acquired, stored, and next distributed?

R2. How can the acquired and accumulated knowledge regarding customer expectations and employees’ expert knowledge of products and components be integrated with data and information from the ERP system in a production process operating according to the mass-customisation strategy?

The first research question (R1) is conceptualised and illustrated in the form of an approach to automatic customised order configuration as an integral part of a diagram of a CPS for a manufacturing process operating according to the mass-customisation strategy (Figure 1). Our approach was developed based on the analysis of the literature on mass customisation with Industry 4.0 and observation of a workflow within a real-life manufacturing company from the automotive industry, which produces personalised mattresses for camping cars.

Therefore, in the literature [14,15], it is stated that the Industry 4.0 concept should take the needs of mass customisation into account [15]. The quick requirements-oriented response needs to be addressed through the use of fully automated and digitalised workflows or processes [16]. Therefore, the challenge lies in how the customer can be involved and integrated into designing and configuring individual products [17]. In the context of Industry 4.0 technologies, this refers to the development of a real-time, up-to-date information flow model [18] through the rapid definition of customer needs and the simplification of the customisation process [19] and decision support system for managers operating within a company in line with the mass-customisation strategy [20,21].

Next, owing to the information acquired from the manufacturing company’s manager, the need to design an algorithm for an automatic customised order configuration was determined.

Finally, each part of the proposed approach is defined and the new algorithm for the automatic customised order configuration is developed (Section 2). R2 is investigated in the form of conducting an analysis of the manufacturing company (Section 3).

2.2. An Overview of Automatic Customised Order Configuration as an Integral Part of a CPS for Manufacturing Processes Operating According to the Mass-Customisation Strategy

Therefore, we propose a new approach to model the algorithm of automatic customised order configuration as a part of a cyber-physical system operating according to the mass-customisation strategy (Figure 1). Two areas have been distinguished in the cyber-physical system, presented in Figure 1.

The first of these, labelled Real World, concerns external sources of tacit knowledge from customers and employees regarding orders and products within manufacturing processes operating according to the mass-customisation strategy.

The second, labelled Cyber World, includes the solution implemented by the enterprise (ERP system) and newly proposed in this article (marked in green) integrating the application of the following methods: natural language processing and rules for converting expert knowledge and collecting the data and information from an ERP system, the use of which will enable the acquisition of ready-made scenarios in the process of accepting an order in the mass-customisation strategy. In Section 4, the research results are discussed, as well as directions for further work.

In the first area of our concept (Real World), the company’s customers configure their customised product using a form on the website and by talking via the phone or via email with an employee of the company. During this process, an employee of the customer relationship management department or sale department writes comments regarding this potential order:

E_not = {e_not1,..., e_notr}, r ∈ N

(1)

These notes can be made using a text editor or be generated from a recording of the conversation with the client using ASR (Automatic Speech Recognition) technology, available via smartphone devices with the Android operating system. To facilitate this process, we suggest using methods from the field of NLP (Natural Language Processing). This will allow words that are similar in meaning to the keywords specified by the trader to be automatically indicated, resulting in the information being formalised and obtained much more quickly. Therefore, it is possible to receive data about the expected product characteristics, which will be stored in the database:

E_pc = {e_pc1,..., e_pci}, i ∈ N

(2)

These are the functional characteristics of the product expected by the customer, such as e_pc1—the expected carrying capacity, e_pc2—the level of resistance to moisture, etc.

Furthermore, owing to the data obtained on the basis of the customer’s order from the website, it will be possible to obtain data regarding the expected product parameters:

E_pp = {e_pp1,..., e_ppk}, k ∈ N

(3)

These are the product parameters expected by the customer, such as e_pp1—weight, e_pp2—dimensions, e_pp3—shape, etc.

In order to build a knowledge base, it is also necessary to receive specialist knowledge from employees regarding the customised product expected by a customer. In the process of the customisation of production—and especially in the aspect of assessing the actual possibilities of producing a specific product configuration—it is necessary to take into account the knowledge of experienced employees. Such knowledge can be obtained through properly prepared questionnaire sheets and then saved in the form of rules that can be stored in the rule knowledge base (“knowledge base”).

Therefore, in our concept, it is assumed that a method based on the rules will be used to formalise this tacit knowledge [22] in the knowledge base:

R_pc = {r_pc1,..., r_pcj}, j ∈ N

(4)

These are the actual functional characteristics of the expected product, such as r_pc1—actual carrying capacity, r_pc2—actual resistance to moisture, etc. In order to store such accumulated tacit knowledge, it is necessary to formalise this knowledge. In our case, only one type of rule is used, i.e., if-type rules, namely condition then results, where the condition verifies only one fact at a time and the result determines the effect on the selected parameter or characteristics of the product. From the point of view of the system discussed, ‘extreme cases’, so-called experiences can be described as, for example, “material A basically combines with material B, but, in practice, the connection is not certain” (negative recommendation), or “material C is very convenient for cutting complicated shapes, in contrast with material D” (positive recommendation for material D). However, as a consequence, the influence, returned as a result of the rule, may take one of four values: excluded, negative, positive, or very positive. We assumed that the rules will be saved in the JSON format, the attributes of which are as follows:

• sourceType—the type of attribute, parameter, or characteristic that is verified in the rule;
• sourceName—the name of this attribute;
• value—the attribute value being checked;
• targetType—the type of target attribute, parameter, or characteristic affected by the value;
• targetName—the name of the target attribute;
• impact—affects the value of the target attribute.

The general format of this rule is shown below:

{

"sourceType":

 "material | accessories | productParam | productCharacteristic | orderParam ",

"sourceName":

 "ex. materials.name",

"value":

 "ex. wool",

"targetType":

 "ex. productParam",

"targetName":

 "ex. products.pickingTime",

"impact":

 "excluded | negative | positive | very positive"

}

In the second area of our concept (cyber world), the two data sources from the ERP system implemented within a manufacturing enterprise are distinguished: data about stock levels, which are a source of information about current production capacity and are necessary to determine whether a specific product configuration can be produced in the time indicated and production capacity, and data about products and orders. The ERP system enables the export of collected data in the form of structured files, such as .csv, .xml, or .json. for this reason, the product characteristics and customer needs can be transferred to the proposed system database. Therefore, the database contains data about real products, imported from the existing ERP system, namely:

R_pp = {r_pp1,..., r_ppl}, l ∈ N

(5)

which refers to the actual product parameters. such as r_pp1—actual weight, r_pp2—actual dimensions, r_pp3—final shape, etc. Moreover, the database includes data about order characteristics:

R_op = {r_op1,..., r_opm}, m ∈ N

(6)

the actual order parameters are r_op1—actual selling price, including profit, r_op2—actual date of completion, etc. The real product is also described by:

P_comp = {p_comp1,..., p_comt}, t ∈ N

(7)

where p_comt is the product’s components.

When the production is small-lot or even individual, it is difficult to clearly indicate how the selection of individual components (P_comp) affects the final characteristics of the final products (R_pc), especially if they do not have discrete values, such as the level of adjustment of the pliability of the mattress to the needs of a person with restless sleep. It is, therefore, necessary for such product characteristics to be defined by experienced experts (area: real world, Figure 1).

Based on the data collected in the database and the knowledge acquired, it is possible to build a knowledge base through the use of the following methods: Natural Language Processing, and rules and stored data from an ERP system. Next, in our approach, the new algorithm will be developed for the automatic completion of orders as an integral part of a cyber-physical system for manufacturing processes operating according to the mass-customisation strategy (so-called decision support for manufacturing processes operating according to the mass-customisation strategy—Figure 1).

2.3. A Real Case Study

In the manufacturing company COWAN Textiles GmBh, a customer who decides to purchase a personalised mattress can fill out a questionnaire regarding his/her preferences on the company’s website. The questionnaire shows the customer’s favourite sleep position, preferred mattress firmness, tendency to perspire, allergies, height, body weight, height of the mattress used so far, and the maximum preferred height of the new mattress. The customer can also send additional comments along with the survey. Then, an employee of the company from the individual customer service department contacts the customer by e-mail or by phone, depending on the customer’s preferences. The employee obtains information about the shape, dimensions, aesthetics of the mattress the client expects, and why they want to change their current mattress, and prepares an offer proposal. This is performed taking into account the technological limitations and characteristics of the raw materials used by the company. At this stage, the availability of raw materials, the price of the personalised product, and the cost of its transport are also determined, which are based on data from the SAP system.

2.4. Collected Data

In the analysed case study, each of the customised products, which are mattresses, is identified by the name [name] and the system code [code] used in the ERP system; the mattress consists of a number of layers; the layers have different functions and are made from different materials and accessories. The materials from which the individual layers are made have their own unique names [name], and it is assumed that each variant of the material, for example, the same foam with different thicknesses, is a separate material. The materials have a precise thickness [thickness] and bale width in which they are supplied [bale width]. Each material must have a specific weight [weight per unit] and cost per unit [cost per unit] of surface area. Therefore, the most important parameters in the so-called “cyber world” (Figure 1) could be defined.

Firstly, R_pp—real product parameters—were identified:

• Dimensions—width [width], length [length], height [height], and the actual physical area [actual area]. This parameter is important if the product does not have a rectangular shape, for example, if it is a trapezoid or triangle. In such a case, the value of this parameter corresponds to the area of the rectangle into which this shape could be entered. The actual surface of the product, including its shape, is [material area];
• Weight [weight]—due to the specificity of the purpose of the product, one of the key parameters of mattresses is their weight. The mass of the product is the sum of the masses of all the layers that make up the product and, possibly, add-ons, such as locks, handles, etc.;
• Time—the total time needed to produce the product [total time] and the picking time [picking time]; all layers, for the purposes of subsequent analyses, are distinguished independently therein;
• Costs—the total cost of production consists of the sum of the costs of all layers of the material [cost of items], total time cost [total time cost], and additional costs [additional costs];
• Prices—the value [min. price] is the price including the actual total cost and the assumed margin, while the [final price] is the final sale price agreed with the customer.

The product components—P_comp—were also distinguished:

• Each of the components of the product is related to the product itself [ProductId] and the material or additive from which it is made [MaterialId];
• In the case of the manufacturing of mattresses, the individual components correspond to their layers. The material layer is described by its actual surface area [actual layer area], the actual material used in making it [actual area of the layer of material], depending on the width of the material supplied, the height of the resulting layer [layer height], and its actual weight [actual weight of the layer], while the dimensions of the layer are defined by the dimensions of the product itself, as in the case of the thickness and weight of the layer, which affect the parameters of the final product;
• On the basis of the actual amount of material consumed per layer, its material cost can be determined [actual cost of the layer].

The data regarding stock levels and production capacity and about products and orders were collected from the ERP system. The example (Figure 2a) shows a single-layer mattress made of Visco sponge (Visco–Auflage), in which Drell Bambo fabric was used to make the cover. Each of the materials that make up the individual layers of the products are considered in terms of their physical properties, such as their weight per 1 m², as well as other considerations, such as the width of the roll of material supplied. A section of the ERP window with the parameters of the material selected is presented in Figure 2b. The entry indicated shows that 1 m² of this material weighs 0.36 kg and is made of 30% bamboo viscose and 70% polyester. In addition, it is supplied in rolls with a width of 2.3 m. This last value is especially important when determining the cost of the product, which actually depends on the dimensions and shape of the product, as it is not always possible to use the entire width of the material, and some of it will become production waste, the purchase cost of which has already been incurred.

From the point of view of the proposed model, the order (R_op) is described by the name [name], which must uniquely identify it, the order date [order date], the price expected by the customer [price expected], and the expected date [order end date]. In practice, some orders are not fulfilled on time; therefore, the actual end date is also specified. The status [status] makes it possible to distinguish between orders that are at the arrangement stage, then transferred for completion, and ultimately completed. As an example, please see Figure 3.

The ERP system collects much more data regarding the order, such as the address for the order and customer data; however, from the point of view of the proposed model in question, it is not relevant. It is assumed that one order may concern one or more products. Orders carried out by the manufacturer in question can be divided into two types—orders placed directly via the online store and orders placed through personal contact with the employees. In the first case, typical products are ordered, which are presented in the form of a catalogue on the website. The products offered there have predetermined parameters and characteristics, and the customer, when making a choice, indicates the expected dimensions and shape of the selected product. The form enabling this selection is presented in Figure 4.

In the developed concept (Figure 1), the second type of order was adopted, where the performance characteristics (E_pc) of the needed mattress were defined based on the needs of customers who use a form that allows their expectations to be defined. The data collected are sent to a company employee who, having the appropriate expert knowledge, is able to propose a product that meets these needs. This is presented in Figure 5.

Next, the parameters in the so-called “real world” (Figure 1) were defined. In the proposed new solution, knowledge of the company’s employees (specialists) about the actual characteristics of the offered products (R_pc) should be acquired. The form enabling the determination of R_pc is presented in Figure 6. These characteristics correspond to those defined by the expectations of the ordering parties.

3. Research Results

3.1. Knowledge Base

Our novel concept (marked in green in Figure 1) was designed and implemented in the form of a WEB application that was launched on the company’s local network (intranet). The application was implemented with the use of PHP version 7.0, in which most of the system’s mechanisms were implemented. The remaining elements were implemented in the form of micro-services in Python and the Flask micro-framework. The application interface could adapt to the size of the screen on which it was used and could be displayed in one of three language versions. A segment of the constructed database that is responsible for storing data from the range presented is shown in Figure 7.

In the first stage, the method NLP (natural language processing) for acquiring the expected product characteristics (E_pc) from employee text notes was implemented. The Python language and the NLTK (natural language Toolkit) library were used for this purpose. The example shows the first step of this procedure, where the record analysed—in this case, the lorem ipsum text filler—was divided into single sentences. In each sentence, the symbols from the so-called stop list, such as punctuation marks, were omitted and it was specified to which parts of speech the remaining character strings could be assigned. In the following step, the mechanism uses the system-defined list of searched keywords in order to find words that are semantically closest to the record analysed. For this purpose, the linguistic corpus wordnet and the wup_similarity () method were applied. By assigning individual keywords to product characteristics, the system could indicate which parts of the analysed record applied to each of them and what expected values were specified. The automatic analysis of the content of the consultant–customer conversation was an element that facilitated the collection of the order data. Additionally, specialist knowledge was acquired from employees in order to collect the data regarding expected product characteristics (E_pc) in Figure 8a and expected product parameters (E_pp) in Figure 8b.

In order to obtain specialist knowledge from employees, the defined (Section 2.1) rules were used. Owing to the acquisition of data from the ERP system and the developed web application, a database for R_op, R_pp, and P_comp was built (Figure 9).

Finally, it was possible to create an algorithm for automatic customised order configuration, the implementation of which would enable support decision-making in a company operating according to mass customisation based on the developed scenario.

3.2. An Algorithm for the Automatic Configuration of an Order as an Integral Part of a Cyber-Physical System

Here, we are looking to answer the following research question: How can a customised product be produced according to its expected parameters and characteristics? We propose the algorithm for the automatic configuration of an order based on the example of a manufacturer of individualised mattresses, intended mainly for the needs of camping vehicles. This algorithm is divided into four stages (Figure 10, Figure 11, Figure 12 and Figure 13). Each of them is presented using pseudo-code and a block diagram.

Assumptions:

Each mattress generally consists of three functional layers: (1) a cover, (2) a comfort layer/padding layer, and (3) a base layer/supporting layer. In the case of layer 2 or 3, they may consist of a greater number of components (e.g., felt backing + highly elastic foam + cotton wool in the case of a comfort layer); however, such a combination is considered as one element with defined parameters and characteristics. Each of the materials used in production must be assigned to one of these functional layers.

Input data:

GET _{(internet form, data entered by the seller, NLP)} product parameters expected by the customer epp₁ (mass), epp₂ (height), epp₃ (price), epp₄ (area)

GET _{(internet form, data entered by the seller, NLP)} expected product characteristics epc₁, epc₂,… epc_i, where i ∈ N

Output data:

pconf = {pfl₁, pfl₂, pfl₃}

where:

pconf—proposed product configuration

pfl₁, pfl₂, pfl₃—product functional layer (pfl₁—cover, pfl₂—comfort layer, pfl₃—base layer)

3.2.1. Stage One

• For each of the functional layers pfl_i, where i ∈ {1,2,3} do

  1.1.

  For each epc_j where j ∈ N do

  1.1.1.

  Find in [knowledgeBaze] the rules, that [targetType] has value „productCharacteristic”, and [targetName] has value epcj.name.

  1.1.2.

  For each found in 1.1.1 rule, remember in [pfl][i].[matchedComponents], [knowledgeBaze][sourceName] and [knowledgeBaze][impact]

3.2.2. Stage Two

2.

For each of the functional layers pfl_i, where i ∈ {1,2,3} do

2.1.

Browse [pfl][i].[matchedComponents] and check if [pfl][i].[matchedComponents] [impact] is not equal „excluded”. If it is, remove all occurrences of it from the list.

2.2.

For each item in [pfli].[matchedComponents] determine the final match value [value] summing up [impact] each of his occurrences, assuming that:

2.2.1.

value „negative” corresponds to the value −1,

2.2.2.

value „positive” corresponds to the value 1,

2.2.3.

value „very positive” corresponds to the value 2.

2.3.

Sort [pfl][i].[matchedComponents] in descending order relative to [value]

3.2.3. Stage Three

3.

If, for each of the functional layers pfl₁…pfl₃ number of components in [matchedComponents] > 0 then

3.1.

Select a product configuration p_conf, where each functional layer contains the component with the largest [value] in the appropriate pfl.

3.2.

For each of the functional layers pfl_i, where i ∈ {1,2,3} do

3.2.1.

Determine the actual layer parameters for the currently selected component:

• if i = 1 _(cover) then
  • rpp_{4 (actual area of the layer)} = (2 * epp_{4 (expected area of the product)}) + (pc _{(product circumference)} * epp_{2 (expected height of the product)})
• else
  • rpp₄ = epp₄

3.2.1.1.

rfl_ip_{1 (actual weight of the layer)} = rpp₄ * rcup_{1 (actual unit mass of a component)}

3.2.1.2.

pfl_ip_{2 (actual height of the layer)} = rcp_{2 (actual height of the component)}

3.2.1.3.

pfl_ip_{3 (actual cost of the layer)} = rpp₄ * rcup_{3 (actual unit cost of a component)}

3.3.

Calculate

3.3.1.

rpp_{1 (actual weight of the product)} = pfl₁p₁ + pfl₂p₁ + pfl₃p₁

3.3.2.

rpp_{2 (actual height of the product)} = pfl₁p₂ + pfl₂p₂ + pfl₃p₂

3.3.3.

rpp_{3 (actual cost of the product)} = pfl₁p₃ + pfl₂p₃ + pfl₃p₃

3.4.

If rpp₁ > epp_{1 (actual mass of the proposed configuration is greater than expected)}

3.4.1.

Find out which of the layers has the most unchecked components and take the next one from [matchedComponents] this layer, which rcup₁ is less than rcup₁ actual one, and at the same time rcp₂ and rcup₃ are not greater than its current values.

3.4.2.

If it hasn’t checked all components yet, go back to 3.2

3.5.

If still rpp₁ > epp₁ then go to point 4

3.6.

If rpp₂ > epp_{2 (actual height of the proposed configuration is greater than expected)} then do the same as in 3.4, take into account rcup₁ and rcup₃

3.7.

If rpp₃ > epp_{3 (actual cost of the proposed configuration is greater than expected)} then do the same as in 3.4, take into account rcup₁ and rcp₂

3.2.4. Stage Four

4.

If (rpp₁ <= epp₁) and (rpp₂ <= epp₂) and (rpp₃ <= epp₃) then

   return p_conf = {pfl₁, pfl₂, pfl₃}

5.

else

   check whether it is possible to change the values of the expected parameters or characteristics

6.

IF answer == YES then

   back to point 1

7.

else

   return empty result

According to our developed approach to automatic customised order configuration as an integral part of a cyber-physical system for a manufacturing process operating according to the mass-customisation strategy (Figure 1), the proposed algorithm may be a part of the decision support system. Therefore, the three types of possible action scenarios in the context of accepting a new order for the production of a customised product can be distinguished (Figure 14).

In the first scenario, the decision-support procedure may propose one out of four solutions (Figure 14a). The algorithm for looking for answers in this type of scenario can be presented in the form of the so-called pseudocode, which is presented in Appendix A. In the scenario presented in Figure 14b, the decision support procedure may propose any number of solutions. Their number depends on the size of the database of previously sold products and the margin of acceptance for individual parameters and characteristics. This means that, at the stage of determining their value, it will be necessary to indicate whether they are very important, moderately important, or not important, i.e., so-called. ‘weights’ In practice, it is assumed that these levels will correspond to the following figures: 0—not important, 1—moderately important, and 2—very important. The algorithm for looking for answers in the second-class scenario can also be presented in the form of pseudo-code, which is presented in Appendix B. In the case of this scenario, as in the previous case, expectations regarding the price and production date should be taken into account before the final presentation of the proposed products. Our developed solution (Figure 10, Figure 11, Figure 12 and Figure 13) applies to the third scenario. In this case, the product configuration, indicated by the contracting authority, was not known beforehand; therefore, the type 1 scenario could not be applied as the expected values of the parameters and characteristics were determined to be very important, which excludes the application of the type 2 scenario.

4. Discussion

It is known that the goal of research should be ensuring customers’ satisfaction [23]. There are many examples of cyber-physical systems in use, but not many examples of the application of mass customisation. Moreover, no solutions have been found concerning knowledge-based, integrated specialist knowledge and data from the information system or decision support regarding the generation of customised orders (

Table 1. Comprising the proposed approach from existing related works.

Reference Paper	Knowledge Base	Applied I4.0 Technologies	Decision Support Regarding Automatic Completion of Customised Orders
[6]	Parameters of a personalised product	Artificial Intelligence tools	Giving the opportunity to move the production process into distribution, which means increasing the possibility of the individualisation of supplied products
[7]	Centralised storage of data and their processing, and communication or transfer of information	Cloud technologies	Allowing for checking availability and the feasibility of customer orders by allowing machines to communicate with each other in order to deliver personalised products as quickly as possible while supporting the customisation of products by centralising management
[8]	Collected data directly from sold products	Artificial Intelligence tools	Forecasting and planning the demand for products
[24]	Centralised storage of data and their processing, and communication or transfer of information	Internet of Things	Various electronic devices are connected to the global network and can exchange data regarding customer preferences and can optimise the production process with each other without human intervention
[25]	Data about the order and the production process	Blockchain and Internet of Things	Decision support mechanism for customers, and data persistence, traceability, transparency, and reliability
[26]	Information on customer specifications and data from the production process	Blockchain	Generating a plan for mass-customised production systems to fulfil the demand, deepening the interoperability of the system
[27]	Data about manufacturers, suppliers, and consumers	Blockchain, big data, Cloud computing, Artificial Intelligence tools, and Internet of Things	Data flow between consumption and production allowing flexibility between manufacturing resources and the mass-customised production processes
[28]	Engineering characteristics definition and customer requirements	Big data and a Kano questionnaire optimisation function	Quick response product configuration system
This article	Customer preferences, integrated specialist knowledge, and data from the ERP (Enterprise Resource Planning) system	ERP systems, Natural Language Processing, rules for converting specialist knowledge, and an algorithm for the automatic completion of orders	Attendance in verifying the feasibility of the order tailored to the needs of the customer in relation to the production capacity of the company and the parameters of the raw materials used for mass-customisation production

).

Table 1 summarises the main features that distinguish the proposed work from existing, related works. It enriches the discussion with a rapid overview of the main outcome of this paper with respect to the state-of-the-art analysed. Going into more detail, Table 1 reports information on related works about the new technologies in the context of the Industry 4.0 concept as applied to mass customisation, the knowledge base as modelled and applied new solutions, and the possibilities of decision support regarding the generation of customised orders.

The concept of Industry 4.0 leads to the transformation of manufacturing enterprises into intelligent production enterprises through the application of I4.0 technologies, such as robotics, cyber-physical systems, smart sensors, IoT, cloud computing, big data, and artificial intelligence [29] in order to increase productivity, efficiency, and economic benefits [30,31]. Article [32] discusses, in detail, the digital twin as an example of a cyber-physical system (CPS), emphasising the importance of artificial intelligence in this concept. Key enabling technologies for the digital twin were presented, such as edge, fog, and 5G, in which physical processes are integrated with computing and network domains. The importance of AI in each technology domain is determined by analysing a set of AI agents at the application and infrastructure levels. It enables the prediction of motion and verifies it experimentally using real data generated by a digital twin for robotic arms. In article [33], by analysing cyber-physical systems from the perspective of artificial intelligence, AI was assigned to computational methods. Distributed artificial intelligence has also been reported to be used in cyber-physical systems for areas such as cooperation and coordination, intelligent agents, and multi-agent systems. It is known that cyber-physical systems are strongly recommended to be applied when manufacturing enterprises operate according to customised production [34]. Therefore, the topic of this article is a response to a research niche found on the basis of the analysis of related research works on cyber-physical systems and decision support in the implementation of production in the mass-customisation strategy (Table 1). This need was also highlighted in [35] in the form of a tool that supports the integration of virtualisation and user preferences for the efficient design process.

This research included the development of an algorithm as part of a cyber-physical system that will enable the integration of customer preferences with the information available in the enterprise’s ERP system and specialist knowledge. For this purpose, a knowledge base for a cyber-physical system was built, rules were developed to generate individual customer orders, and a cyber-physical system was applied for an exemplary real mattress-manufacturing company.

In our case study, the following further directions for our work are indicated. It would be possible to create another group of products already delivered to the customer, to which changes could be introduced. These changes would provide information for more satisfactory and optimal implementation of similar future orders. One of them would include articles where configuration changes could be made, taking into account possible customer complaints. The second one would concern products, the configuration of which could be changed in terms of, for example, the technological efficiency of the assembly by analysing the perceptions of employees participating in the production process. The application can also be extended with other groups of personalised products that COWAN Textiles GmbH offers to its customers, especially the seats of motorhomes.

5. Conclusions and Managerial Implications

In this paper, we developed an algorithm for the automatic configuration of an order as an integral part of a cyber-physical system in a manufacturing process operating according to the mass-customisation strategy. We elaborated a knowledge base containing a formal record of the knowledge acquired regarding customer preferences owing to the use of rules and data from the ERP system regarding the production capabilities implemented in a manufacturing enterprise. Moreover, the decision rules for the scenario of generating possible orders for a client were established. Finally, we verified our system in a real company that manufactures mattresses.

From a managerial point of view, the proposed system supports decision-making regarding accepting a new order to produce a customised product. It allows, based on previously collected and converted data, and information about customer expectations and their integration with data from the ERP system, for automatic order selection in accordance with the individual needs of the customer. The application of the proposed solution to a manufacturing enterprise operating according to the mass-customisation strategy will increase customer satisfaction, but it is still not clear whether the implementation of the new technology will be profitable. According to [36], in our further research, an analysis of the profitability of the introduction of assistive technologies, such as digital instructions, should be conducted. However, owing to our research results, the managers will receive a tool that supports a faster response to customer needs and they can still maintain their right to make decisions in face-to-face interactions with customers.