Currently engineers who design and analyze chemical engineering schemes (CES), must be able to create and share databases of developed schemes. Such schemes include their graphical representation and calculated results. However, now most of these schemes exist in some graphical format of the drawing software (AutoCad), which is not integrated, or poorly integrated (via VBA) with a performance of CES calculations. Specialized commercial systems provide such an opportunity based on proprietary formats, but most of them are not designed for computer synthesis based on the non-graphic description of the technological schemes.

Problem formulation

In our opinion, specialized public computer formats are necessary for the effective information exchange between the developers of CES (both public and commercial systems). In this way, the scheme should be objectively presented, regardless of the style and the technical taste of the developer. So, the scheme must be unified. In other words, the scheme, which was made by different people using textual description or non-unified drawing, will look the same, and computer-generated descriptions will be also identical. This identity will not only improve communication between developers but will also make it possible to maintain databases of possible CES configurations.

Analysis of research and publications

Unified ways of presenting different information (graphic, text, numerical) in XML and its derivatives (GraphML) are widely used in modern programs. However, there are no unified ways of CES representation, primarily in configuration aspect. Different experts using a textual description of process steps and their sequence, including recycles and bypasses will draw different schemes realizing the same description. Coding such schemes in the XML format will give identical files, and the more complex the scheme is, the more options of the text description are. Thus, in programs that use XML to describe the calculation schemes (for example, Xcos SciLab), CES parameters will be calculated several times with the same result. Another significant problem is a lack of a single CES representation during expert software development of technology schemes synthesis. The program cannot distinguish between different versions of the same description of the some scheme and different schemes which realize the same problem without the unified approach to the scheme description.

A common rule of advanced calculations, including calculations of chemical and engineering industries, is “withholding” of computing procedures through the provision of services by any graphical interface. Also, universal program complexes (Hysys, ChemCAD, etc.) for calculating chemical processes do not provide information about saving the structure of CES. Such closeness raises a number of organizational and technical problems.

After examining a number of publications, initiatives aimed at the unification of the exchange and presentation of CES topology were identified. In particular, it was XMLPlant [1], which used XML. An XML document consists of text characters and is suitable for reading by human and computer processing in order to storage structures. Another program is CAPEOpen [2], which is a standard of data exchange between programs of chemical technology modeling. These initiatives have not got recognition among manufacturers of software systems. There are some software products on the market that allow you to calculate the flows of CES, but all these programs strictly limit data exchange.

Unification of CES presentation

We propose a unified way of CES drawing (Figure 1). It allows creating the table (Table 1) and encoded text descriptions of engineering schemes. Then obtained codes are translated into XML format and are used directly in programs using languages of general or special purpose.

Table 1. Table CES description

It was an example of a 2-stage reverse osmosis (RO) installation. It shows that there is an opportunity to perform simple text encoding of lines with equipment (capital letters) and without them (lower case letters). Recycling  is described using reverse (in alphabetical order) sequence of letters:

1:aBCd;2:ba;3:ca;4:b2c3.

Upper horizontal processing line (level 1) is the main flow (from the initial substances to the key product). Choice of the main flow is based on different factors, for example, productivity, a number of devices (if the line with the maximum flow rate is not obvious), functionality (in processes of water purification). Choice of the main factor is agreed on beforehand. The sequence of the next levels depends upon the order of the following levels. Lines with equipment are usually higher than lines without it (recycling and bypass).

Proposed way of CES representing does not always use the space of the drawing sheet effectively, but it allows you to get the same drawings and code by experts who use text description or non-unified scheme. In addition, formal scheme control is improved (number of levels for the technology and text encoding). That is important in teaching and for designing critical engineering areas.

Figure 1 – Two-stage RO installation scheme with partial concentrate recycling (Excel editor).

XML based configuration of CES

There is software for a unified representation of CES. Those programs can use CES for analyzing and choosing the best configuration after performing some calculations.

It is recommended to use yEd as the best editor to realize such an idea. This editor has worked well on all popular platforms and has algorithms for automatic visualization of the arbitrary graph structure. In addition, the editor stores information in GraphML format (based on XML), and that allows performing the calculation of a scheme using information about its configuration.

For example, Figure 2 shows the scheme of RO installation, that was created in yEd.

The used tool gives the opportunity not only to set the topology of the water consumption scheme but also to set the parameters of the devices and streams that will be saved along with the configuration of the scheme (in XML file). In the illustrated example such parameters are flowrate and composition of the input streams for the 1st and the 2nd stages of water treatment.

Figure 2 – Two-stage RO installation scheme with partial concentrate recycling (yEd editor).

Shortened recording of the scheme (XML-file fragment describes one of the graph edges) is as follows:

<edge id = “e0″ source = “n0″ target = “n2″>

<data key = “d10″>

<y: ArcEdge>

<y: Path sx = “0.0″ sy = “0.0″ tx = “0.0″ ty = “0.0″>

<y: Point x = “85.0″ y = “112.0″ />

</ y: Path>

<y: LineStyle color = “# 000000″ type = “line” width = “1.0″ />

<y: Arrows source = “none” target = “standard” />

<y: Arc height = “0.0″ ratio = “0.0″ type = “fixedRatio” />

</ y: ArcEdge>

</ data>

</ edge>

It could be seen that topology and visualization data are partially mixed. It is not essential in further scheme calculation using programs because the structure of XML format allows you to choose data that you need for a particular application.

XSLT technology can be used in order to create automatically a unique index which will identify the scheme. XSLT (eXtensible Stylesheet Language Transformations) – is a language of XML-documents transformation. XSLT specification is a part of XSL and is recommended by W3C.

Graph format based description      

The proposed approach to coding CES has a disadvantage. It is redundant information for the specialist in the subject area. On the other hand, table-based representation does not use paper space efficiently when drawing.

Currently used text formats for coding graphs don’t have the first disadvantage. Special programs (for example, free GraphViz and yEd) have a variety of options for displaying a predetermined configuration. That optimizes the space of the sheet to display the scheme and overcomes the disadvantages of a table view.

The are two graph formats: TGF (Trivial Graph Format) and DOT (extension .gv or .dot). DOT format (GraphViz) is not suitable for yEd, but you can convert DOT into GraphML to run it in yEd.

Explicit representation of a mixer and a separator in a directed graph is used in order to represent a single-stage scheme of RO installation in DOT-format (undirected – graph):

digraph osmos1 {

“mixer” -> “membrane” -> “permeate tank”;

“membrane” -> “separator” -> “mixer”; // concentrate recycling

“Separator” -> “sewage tank”;

}

The sequence of relation descriptions in a graph is not important. For example, if you put the line with concentrate recycling at the third line and not at the second one, then the graph will be displayed by the identical manner according to the display style. However, textual representations of the two options (after converting a text file into a string) will be different when using computer functions of line comparison. Consequently, the proposed unification of CES representation (sequence of scheme levels) remains relevant for the described formats, since it provides an unambiguous identification of schemes with horizontal encoding.

Conclusions

Thus, the use of the proposed approach is useful for synthesis, calculations and analysis of any chemical and engineering schemes. The format of technological schemes based on tables, graph formats, XML (GraphML) and related information technologies will reliably verify the schemes and provide access to their extensive reciprocal exchange, including commercial basis.

Further calculations using data on the scheme configuration and its parameters can be implemented on general-purpose languages (Python, VBA), as well as in specialized mathematical programs (SciLab, MatLab).

Such types of complexity were characteristic of the pre-computer era, and at the present time, neither a large number of elements in the scheme nor the presence of recyclings is a serious obstacle to the calculation of the parameters of technological schemes. Therefore, consider these “problems” in general and on specific examples.

Apparatus as a function
The most actively functional approach [1] is used in modern programming languages (Erlang, Haskell, Clojure, Scala…). A characteristic feature of this approach is the use of functions as “black boxes”. This means that the internal logic of the functions is hidden from the environment, and the interaction between them is carried out through their inputs and outputs. That is, the outputs of some functions (the results of their calculation) are inputs (parameters) of other functions.
In the same sense, we can also consider apparatuses in chemical-technological schemes, since this corresponds to the actual state in them. Each device interacts with the other only by taking and giving out streams of matter passing through them.

The peculiarity of the proposed solution is the use of the non-functional programming language VBA in the programming environment of Excel which is reactive [2] in the sense that any changes in the values in the cells result in automatic recalculation of all dependent cells. In this environment, any VBA functions interact with each other through the I / O cells of the Excel sheet and in no other way. In fact, the ability to define user-defined functions in VBA is sufficient to implement the proposed approach. That is, any programming language that supports the definition of functions can be used. The second important feature of the proposed programming environment is the possibility of implementing iterative calculations, which avoids programming cycles in the VBA language by hand.

Another important feature of the calculation’s organization on the Excel sheet is its visibility. Excel sheet allows displaying the technological scheme in the usual way on it, while in the place of the displayed devices will be placed calls of the corresponding functions. The result of such a calculations’ organization will be a “live” scheme where changing the input parameters leads to a change in all output parameters.

Calculation of the reverse osmosis installation
As it was mentioned above, all apparatuses in chemical technology are divided into three types (and their combinations) from the point of view of calculations based on the material balance [3].
To illustrate the basics of the proposed approach presented the reverse osmosis (RO) unit with concentrate recycling that is selected since it consists of just these three variants and one of their combinations. The membrane is a reactor and a splitter at the same time: water desalination occurs in it and the input stream is divided into two: permeate and concentrate. The splitter separates the concentrate stream into recycling and sewage, and the mixer mixes the recycle stream with the raw water and feeds this mixture to the inlet of the membrane.

Figure 1. Scheme of the single-stage RO unit with concentrate recycling

The main objective of this calculation is to determine the allowable degree of recycling (Xr), which is defined as the ratio of the recycling flow (Gr) to the concentrate flow after the membrane (Gk). It is known that exceeding the permissible level of recycling leads to the formation of deposits on the surface of the membrane. The parameters Xr and Gk are the input parameters of the splitter and the corresponding function (fig. 2), and the recycle flow and sewage flow rates (Gs) are the output parameters (the results of the function calculation). Obviously, TDS (total dissolved solids) of water does not change on the splitter and therefore this parameter does not use for this function.

Equations for the function of the mixer are derived from the material balances of water flows and the flow rate of dissolved solids per hour. The input parameters are the water’s flow rate and TDS of the recycling water, and the flow rate of raw water and water at the outlet are the output parameters for the mixer (Gv). The output parameters are the flow of the raw water (Gi) and TDS of water at the outlet from the mixer.
It should be noted that physically the input parameter Gi became the output parameter for the function due to its dependence on the recycle flow rate since the flow rate at the outlet from the mixer is a constant and is determined by the performance of the membrane. For example, a reduction of the recycling rate (Xr) results in a reduction of the recycle flow, which will lead to an increase in the flow of raw water while the water flow rate at the outlet from the mixer will remain constant.

When organizing the calculations of the membrane, all the principles of material balances for water flow rates and water’s TDS (in form G*TDS) are taken into account. In addition, the parameters of the membrane itself are used – the permeate output (Xp) and the water desalination degree (Xd).

The functions developed in VBA are shown in Figure 2.

 Figure 2. The  VBA functions

Note that to get access to user-defined functions via the standard menu, one have to place them in a module that is not present in the spreadsheet by default.

Assembling the functions in Excel

The assembly of functions on the Excel sheet with the recycling scheme can begin with any element, but we started with the membrane, continued with the separator and ended with a mixer (fig. 3).  Note that after displaying one calculation result for the used functions with 2 or more results, it is necessary to select (from left to right) the number of cells equal to the number of return values, then press the F2 key, and then simultaneously press the Ctrl, Shift and Enter keys.

Figure 3. Embedding a custom function in Excel

To start the calculation of the membrane, an initial approximation is necessary for the value of the water inlet TDS. At the end of the assembly, this value will be replaced by the calculated value from the function of the mixer. To use the iterative capabilities in Excel, one must enable the appropriate setting in the calculation parameters.

Figure 4.  Results of calculation of single-stage installation

An interesting feature of such calculations in Excel is the loss of calculated values after opening a saved file (fig. 5). 

 Figure 5. The loss of calculated values after opening a saved file

To fix this error, one must delete one of the functions (not always obvious which one), and undo this action (fig.6 and fig.7).

Figure 6.  Deleting one of the functions

Figure 7. Undoing delete action

To check the performed calculations, it is necessary to use the “big black box” principle, which allows estimating the flow rate at the input and output from the installation without taking into account the processes occurring inside the scheme. If these values do not match, then when making calculations, errors are made. The discrepancy between the input and output of the unit is less than 0.1% (1 g/hour from 540 g/hour), which indicates the correctness of the calculations performed.

Extending the calculation to a 2-stage RO scheme with recycling

Extending the calculation to a 2-stage scheme involves developing the function of the 2nd mixer (fig. 2) and using the same splitter and membrane functions, but with different input parameters (fig. 8). Obviously, for the second membrane, the input parameters will be the output parameters of the first one. The second mixer is “normal” in the sense that the parameters of its inputs (recycles after the 1st and 2nd splitter) are known, and the output parameters are calculated .

 Figure 8. Embedding additional custom functions

 At the end of the calculation’s assembly, it is necessary to change the input parameters of the first mixer, the second input of which takes a mixture of two recycles instead of one (fig. 9). It should be noted that in practice the second recycle is usually not used in this way.

Figure 9.  Changing the input parameters of the first mixer

The discrepancy between the input and output of the unit is less than 1% (3 g/hour from 350 g/hour), which indicates the correctness of the calculations performed.

It seems natural that by expanded the calculation scheme once it is possible to do this as many times as necessary with the same functions or with others.

Conclusions

A simple technique for calculating the parameters of technological processes of any complexity level is presented. For its implementation, the elementary knowledge of VBA language is required at the level of the organization of calculations of three types of apparatus and their combinations in user functions. The calculation of schemes with recycling flows requires the iteration possibility of Excel.
Such calculations lead to some easily solved problems associated with loss of calculation values after opening a saved file. At the same time, visualization of connections between functions on the Excel sheet, simplifies the organization of calculations, especially for beginners.
An important feature of the proposed approach is the possibility of independent creation of functions library by different developers. This is due to the functional approach applied with VBA in Excel environment, which ensures the reactivity of the calculations.

A common use of the LINEST function

To solve this problem, the built-in function “ЛИНЕЙН” (LINEST) is used, the possibilities of which are given in Excel’s help or in books [1]. Usually, this function is used to obtain linear dependencies from one or more parameters. To use it for nonlinear dependencies, such initial data are preliminarily linearized by logarithm. An example in chemistry is the processing of kinetic data to obtain values of the order of reaction, rate constants, and activation energy.

However, not always linearization using logarithm leads to the desired result and therefore the standard use of the LINEST function is not applicable. The solution is the “deception” of this function, in which a non-linear relationship is organized on a sheet of the spreadsheet in the required format. It should be noted that the choice of the initial non-linear dependence (power series, exponential or any other) is made by the user. That is, the accuracy of the obtained approximation will depend not on the LINEST function, but on the chosen nonlinear dependence format.

Let’s consider the given approach on a concrete example of approximation of dependence of function on two and three parameters: degree of oxidation in a chemical reaction from the concentration of a reagent and time of contact. We first estimate the accuracy of a simple linear approximation for these data (fig. 1).

Figure 1. Data processing with LINEST function

The results of the study were obtained on the basis of a previously designed experiment [2] on the effect of temperature and concentration on the degree of conversion. Naturally, any set of data can be approximated provided that such data is sufficient to determine the unknown regression coefficients.

The results of a simple linear approximation (fig. 1) indicate a complete coincidence of calculations based on natural (regression coefficients bi) and coded (regression coefficients ai) parameters. Relative deviations are up to 3 percent in the main. However, the 2nd point with a slight absolute deviation gives 27% of the relative deviation.

“Cheating” the Function

The basis of the proposed approach is the deception of the LINEST function, in which a non-linear dependency is formatted for use with a function that processes linear dependencies. Figure 2 shows the results of calculations for a nonlinear second-order model with natural parameters. The level of relative error is 0.01-0.04%.

Figure 2. “Cheating”  the LINEST function

It should be noted that the LINEST function returns the regression coefficients in the reverse order. To obtain these coefficients, one need to use the procedure “select, F2, Ctrl-Shift-Enter” [3]. Interestingly, the Calc program in LibreOffice displays the results of the function with several results without additional user actions (fig. 3)

Figure 3. “Cheating”  the LINEST function in Libre Office without ”select, F2, Ctrl-Shift-Enter” action

Application to a function of three variables

Let us consider the results of approximation for a variant with three parameters. The same trend with respect to the accuracy of simple linear (fig. 4) and non-linear approximations (fig. 5) is retained for this variant.

Figure 4 - The usual application of the function

Figure 5. The unusual application of the function

It’s obvious that the inaccuracy of simple approximation increases and the accuracy of nonlinear approximation remains at a very good level.

Conclusions

The presented approach to the treatment of experimental data has been used for many years at the Authors’ Department in KPI. In addition to solving actual problems in the processing of experimental data, this approach is of pedagogical interest as an example of a non-standard solution based on the well-known. It should be noted that neither the Excel Help nor the excellent book by John Walkenbach considers such application of the LINEST function for data processing.

In this regard, the approach in the study of almost any subject looks promising through the creation of expert systems and models in the field under study. It is quite obvious that when developing an expert system, the developer must understand this subject, and in the development process the level of understanding of this area inevitably increases due to interaction with the already operating part of the system. The previous article [2] is devoted to mathematical modeling, a simple technique for developing expert systems in the visual programming language Drakon [3] is considered in this work.

Quick start with the Drakon – 1 step

It is known that the first steps in any new business are the most difficult. The designing of expert systems is quite complex, although in fact, the development of such systems is based on the ability of any person to ask questions and answer them (although not always correctly). Teaching this subject for more than 5 years in the KPI, I used other approaches to the creation of expert systems (CLIPS, decision tables), but the solution based on the Drakon seems to be the best for beginners because of the rapid entry into the topic based on just two steps.

The first step is based on using the Drakon Editor [4] to solve a simple task (displaying a string on the monitor on condition its length is permissible), an example of which is shown on the program’s website, and the code is generated for the Python language (fig. 1). After that, the code is run offline (in IDLE for example) or online (repl.it). At this stage, the whole cycle of work with the Drakon is mastered: drawing a block of the circuit, automatically checking it, and generating the code in the chosen programming language, executing the code in the appropriate environment.

Figure 1. The example from the Drakon Editor website

Figure 2. The generated code on Python 3.6

Figure 3. Running the code on the site repl.it

To understand the features of working with Drakon, my students are given the task of generating code for the same task for JavaScript language, and to launch the code in the browser. At the same time, the block diagram is not altered, but only 2 functions in the diagram blocks are changed: the output of data to the monitor (from print to alert) and the function of the string’s length (from len(string) to string.length).

Figure 4. Running the JavaScript code in the browser

The automatic generation of program code without changing the visual scheme for almost any procedural languages makes the use of Drakon so universal. Of course, within of solving such a simple task, one can not evaluate all the possibilities of the Drakon, but it allows one to quickly learn the procedure for using the Drakon with different programming languages.

The expert system of a Trip by bus – 2 step

In the book of the Drakon language author’s [5], there is a trip algorithm already on 21 page that is not associated with a specific programming language (fig. 5). It is interesting that this example already gives all the necessary information about the creation of any small expert system. Adaptation of this algorithm for Python (fig. 6, 7) requires only two input and output operators and allows estimating the difference between the generated code (see the figure) and the source diagram (see the figure) in sens of plainness for a reader of an algorithm.

Figure 5. The algorithm of a trip by bus

Figure 6.  The adaptation of the algorithm, part 1

Figure 7. The adaptation of the algorithm, part 2

The difference between the second and the first diagrams is the appearance of a structure for the organization of the cycle, which in combination with the conditional operator (presents in both diagrams) is a necessary condition for the development of an expert system.

Figure 8. The generated code on Python 2.7

Figure 9. Running the code in IDLE

Running the generated program in Python allows one to evaluate the correctness of the algorithm from the perspective of an expert, which is any user of public transport.

It is important that for the large expert systems the approach remains the same, but the final project will consist of several diagrams and the functions generated from them that interact with each other.

Conclusions

In this article is presented an approach to rapid learning in the development of expert systems based on the visual Drakon language in the programming environment of the Drakon Editor, as well as the launch of automatically generated code in the Python language for interactive evaluation of the developed algorithm. This approach consists of only two steps (developed diagrams on the Drakon) and allows students to design their own expert systems for the studied subjects. The ability to adapt the diagrams on the Drakon to a variety of programming languages without changing the visual part (actually the algorithm) makes it possible to widely use this approach for various real-world problems, for example in web programming using JavaScript, Java, and PHP languages.

Essentially, computer’s training by creating an expert system is very different from human learning, since an incomplete description of the problem or its inadequate solution almost immediately (quite often) becomes apparent in the testing of the program even for a not very competent developer. Of course, for the final evaluation of the developed system, an expert will be needed: a teacher at the training stage or the customer after study.