Industrial extruder, process and method

Abstract

An industrial smart extruder (1) and extrusion method are proposed, in which the industrial intelligent extruder (1) functions as a conveying device that uniformly extrudes a viscous mass from a solid through a forming orifice under high pressure and temperature according to the Archimedes screw operating principle, and the material is processed by the extruder (1) through hot extrusion, cold extrusion, warm extrusion, friction extrusion, or microextrusion, and the material extruded by the extruder (1) includes at least food, metal, polymer, ceramic, concrete, or modeling clay. The industrial intelligent extruder (1) has a smart device with an ML or AI-based core engine that automatically controls, steers, and / or optimizes the operation of the extruder (1).

Description

[Technical Field]

[0001] The present invention relates to the field of extruder systems and processes. In particular, the present invention relates to an industrial extruder system having a feeder, an extruder, a die, a collection means, and an extruder control. The feeder supplies plastically deformable and / or viscous input materials to the extruder, which continuously presses the input materials into and out of the die to form an output material as an extrudate according to the Archimedes screw principle of operation. This process is called extrusion. The present invention also relates to both processing and compounding extruders. Processing extruders can be used for shaping (typically piston-type and single-screw extruders), while compounding extruders are used for chemical and physical modification (reaction, mixing, degassing, etc.) of materials (co-rotating, close-contact twin-screw extruders, busco-kneaders, etc.). The extrusion process of the present invention can involve hot extrusion, cold extrusion, warm extrusion, friction extrusion, and / or microextrusion. The extruded materials can include, inter alia, food, metal, polymer, ceramic, concrete, or modeling clay. Finally, the present invention relates to technical solutions that address the possibilities and challenges of applying ML-based techniques to hybrid food processing operations, in particular solutions that address the potential of physics-based ML modeling techniques for food processing applications and appropriate optimization and / or automation architectures. Generally, the present invention provides a valuable contribution to advancing ML-based techniques for food processing applications. [Background technology]

[0002] It is generally known that industrial extrusion machines squeeze input materials, including raw materials, and apply pressure until they exit the extruder through a die. While there are many different variations in extruder equipment, most commonly they include a hopper for holding and feeding the raw materials and a hydraulically or mechanically driven means for applying pressure. Most extruders also include a die for shaping the final extruded product (extrudate). The means for applying pressure typically includes a single or twin screw auger driven by an electric motor or a hydraulically driven ram. Screws commonly used in extruders include single-flight metering screws, single-barrier screws, double-barrier screws, variable-pitch flights, multi-start flights, slotted flights, and two-stage screws such as vented extruders.

[0003] Extruders are commonly used in polymer processing. Extrusion, a technological process in which molten polymers are forced through a die, is used to produce components of fixed cross-section, such as tubes and rods. Polymers are substances or materials composed of very large molecules, or macromolecules, composed of many repeating subunits. Due to their wide range of properties, both synthetic and natural polymers play essential and ubiquitous roles in everyday life. Polymers range from familiar synthetic plastics such as polystyrene to natural biopolymers, such as DNA and proteins, that underlie biological structure and function. Both natural and synthetic polymers can be produced by the polymerization of many small molecules known as monomers. This results in larger molecular weights than small molecule compounds, resulting in unique physical properties, including toughness, high elasticity, viscoelasticity, and a tendency to form amorphous and semi-crystalline structures rather than crystalline structures. Food polymers are polymers derived from edible plants, animals, and microorganisms that can be used in food systems, including proteins, polysaccharides, and peptides. Generally, food polymers can be classified into three groups based on their source: (1) plant-derived food polymers such as starch, dietary fiber, and grain proteins; (2) animal-derived food polymers such as meal proteins; and (3) microbial-derived food polymers such as fungal polysaccharides. Oils and / or lipids derived from plants and animals can also be considered food polymers, despite their relatively small molecular weights. Thus, food polymers occupy a major area among natural polymers and play important roles in food structure, food functionality, food processing, and shelf life.

[0004] Extruders, in particular, are popular in food processing because they create a rapid, continuous process that can be used in the food industry to produce many foods, such as snacks, breakfast cereals, pelleted products, pet food, and pregelatinized flour. These systems incorporate multiple unit operations, such as mixing, kneading, cooking, shaping, and cutting, into a single piece of equipment. This results in a relatively simple process with high efficiency and low cost compared to other processing methods. As such, extrusion is a widely used processing technique in the food industry and has a wide range of applications. During the extrusion process, the input material, consisting of food ingredients, is subjected to high shear, temperature, and pressure for a short period of time. This helps convert the ingredients from a solid powder to a molten state inside the extruder. The molten ingredients are then forced through a die at the end of the extruder into the atmosphere. The melt exiting the extruder experiences a sudden drop in pressure, resulting in rapid expansion and a drop in temperature, which aids in its conversion into a cooked product. A schematic diagram of the conversion process is again shown in Figure 1. The final quality and texture of the extruded product depend on various factors, including the ingredient mixture and its properties, the extrusion conditions, and post-processing conditions. Because extrusion is so flexible, it has found a wide variety of applications in the food industry, some of which can be seen in Table 1. Compared to other common food processes, extrusion has great advantages due to its applicability to a variety of food processes, flexibility, cost-saving effects, high production rates, and high-quality products.

[0005] [Table 1]

[0006] Figure 2 shows a high-level block diagram of what a food extrusion production line might include. The process begins with characterizing and receiving the raw materials. The raw materials used are critical to the consistency of the product at the end of the processing line. The raw materials then undergo blending and / or preconditioning (see Figure 30), which can be done using equipment such as ribbon blenders or preconditioners to ensure uniformity as they enter the extruder. However, blending and preconditioning are optional for certain products. This is followed by extrusion (see Figure 31), which is the primary cooking step where the raw materials are cooked and transformed into a formed product. Post-extrusion processing operations include cutting the exiting product to the appropriate size, drying the product to the desired moisture content, and seasoning or coating to impart the desired flavor and taste to the product before proceeding to packaging. In addition to these primary processing sequences, there may be additional steps depending on the type of product being produced and its intended use.

[0007] Extrusion processes are typically defined as either "hot," "warm," or "cold" extrusion. Hot extrusion is performed at elevated temperatures to prevent work hardening of the material and make it easier to push, especially as it passes through a die. Warm extrusion is performed above room temperature but below the recrystallization temperature of the material. It is often used to balance the required force, ductility, and final extrusion properties. Cold extrusion is performed at or near room temperature. Cold extrusion has several advantages over hot extrusion, including a lack of oxidation of the material, closer tolerances, and better surface finish.

[0008] As mentioned above, extruders are widely used in the food industry. They are used for simple forming applications such as pasta, or for more complex operations that involve significant modification of the extruded material. The extrusion feed material, for example, can contain a variety of solid and liquid components that result in ill-defined transformation reactions and rheological / viscosity fields.

[0009] The main components of an extruder include the feeder, barrel, screw(s), and die, but more components can be added to increase product versatility. The feeder is used to continuously feed the mixture into the extruder at a constant rate to ensure consistency. Feeder feeders often deliver materials by weight or volume, and more than one feeder can be used at a time for different ingredients. The barrel houses the screw or set of screws. The barrel is often jacketed for heating and cooling. Heating can also be achieved by installing an electric heating unit on the barrel or by steam. The barrel's inner lining is often smooth in twin-screw systems, but may be grooved or fluted in single-screw systems. The barrel may also have various injection or additional feed ports along its length. While injection ports can be used for water or other liquid ingredients, additional feed ports can force additional powdered ingredients through the side of the barrel, bypassing a large section of the extruder and resulting in less cooking. The role of the screw(s) is to impart shear to the raw material mixture, helping to force the combined dough through the die and out of the extruder. The screw is also responsible for the pressure buildup at the end of the extruder and additional mixing of the ingredients. The die serves to hold the material in the screw, allowing time for the screw to impart shear energy to the sample. The die also controls the final shape of the product, which can be very diverse. The final critical component of the extruder is the motor, which provides the energy needed to rotate the screw. Other optional components can include, for example, a preconditioner, which can be used to pre-hydrate the raw material mixture and, in some cases, pre-cook the material before feeding it into the extruder. A die cutter is also frequently used to help cut the final extrudate as it emerges from the die.

[0010] There are two main types of extruders: single-screw and twin-screw (co-rotating and counter-rotating). These come with a wide range of screw diameters and length-to-diameter (LID) ratios, with or without steam preconditioning. Given the limitations of current processing know-how, optimally selecting and designing an extruder for a specific application is not always easy. The various known extrusion processes can be broken down based on functionality and operating parameters, addressing different processing needs, and / or differing by a logical approach to the selection or design process. Fundamental is understanding the interplay between material properties, rheology, conversion kinetics, and system design and operating parameters.

[0011] For example, a discussion of the prior art can begin with the use of single-screw extruders to plasticize / melt synthetic polymers, since most of the development of extrusion theory is based on this application. Comparing plastics and food extrusion can then be used to learn from and properly analyze single-screw food extrusion processes. Co-rotating and counter-rotating twin-screw extruders are also found in various forms and can be compared to single-screw extruders and each other.

[0012] Food extruders utilize both thermal and mechanical energy. Understanding energy consumption and input requirements is crucial for improved performance and economical system design. Water is a common ingredient in almost all food extrusions. A portion of the water can be applied in the form of steam, thus supplementing the extruder's total energy requirements. This option can significantly impact extruder selection, design, and performance, as well as product characteristics. Combined extrusion cooking and shaping of unexpanded products is a common process used to produce snacks, cereals, and other products. In such operations, accelerating cooling before the final shaping step is important. This is typically achieved by evaporating a portion of the liquid water within the extruder or during an aeration stage between the cooker and the shaping extruder. When extrusion products are developed on a low-volume production basis, scale-up is part of the overall extruder selection and design process. Understanding the scale-up process is important for identifying its critical factors, limitations, and complementary process options. Finally, it should be noted that extruder design parameters and operating variables generally interact and thus affect extruder performance and product characteristics. Therefore, more than one system design and operating option may be available to produce the same product. The optimal choice is likely to be governed by availability, flexibility, and economics.

[0013] Modeling the extrusion behavior of plastics and food

[0014] Non-food polymers can be divided into three main groups: thermoplastics, thermosets, and elastomers. Thermoplastic materials soften when heated and harden when cooled. Generally, their chemical composition does not change during extrusion. Thermosets undergo a cross-linking reaction when the temperature rises above a certain limit. This results in the formation of a three-dimensional network that remains intact even when the temperature drops. Because the cross-linking reaction is irreversible, thermosets cannot be recycled as thermoplastics. Elastomers, or rubbers, are materials that undergo significant deformation when subjected to shear forces but return to their original shape when the force is removed.

[0015] In the prior art, development of polymer extrusion technology has focused primarily on the use of thermoplastic extruders. In most cases, only a single polymer feedstock is extruded. The process involves conveying to a die, plasticizing (melting), and melt metering.

[0016] Two types of polymerization reactions, "addition" and "condensation," produce plastic polymers. In the first type, individual monomers, such as ethylene, add directly to each other without changing composition, forming long-chain molecules. In the second type, two or more monomers react with each other, eliminating a small portion of them (usually water) to form linear molecules. The reacting monomers in this reaction type usually have two functional groups, which can react with each other sequentially. Examples of this type of polymer are polyamide (nylon) and polyester. All linear or slightly branched polymers produced by addition or condensation reactions are thermoplastic; that is, they can be repeatedly softened at high temperatures and solidified again by cooling. No chemical changes occur during these heat treatment processes.

[0017] Extruders are typically used to convey, melt, and mix solid polymers and form final product shapes such as sheets, pipes, rods, etc. They are also used to produce plastic raw materials such as compounds and pellets for further processing operations. In addition to continuous operation, extruders can also be used semi-continuously, in the case of injection molding machines, by reciprocating the screw forward once the mold-fill cycle begins.

[0018] Extruded food ingredients are composed mostly of carbohydrate (e.g., starch, fiber, sugars) and protein biopolymers. Other ingredients, such as water, fats, minerals, and vitamins, are also included in the feed mixture. The structure of biopolymers is not well defined and varies depending on the source. In food extrusion, the process typically involves feeding a mixture of solid and liquid ingredients. These ingredients react in complex, ill-defined processes that often involve irreversible changes in physical and chemical structure. Inside the extruder, the ingredients form a fluid-like substance resulting from either: (1) a low-temperature dough formation process in the presence of sufficient fluid (water and / or oil). Water is always present and acts as a plasticizer, lowering the glass transition and melting temperatures. Components of the formed dough can undergo structural changes if sufficient energy is provided. For example, starch granules undergo a gelatinization / melting process above a certain temperature range. The latter can result in a significant increase in viscosity due to the increased water absorption capacity of gelatinized starch; (2) at low water / oil concentrations, the softening and melting process of biopolymers (carbohydrates and proteins) can be achieved by providing appropriate energy. This can take the form of solid friction and mechanical energy, steam energy, and / or viscous heat dissipation through the barrel. Usually, the food material in this case also undergoes irreversible physical and chemical changes.

[0019] In most food extrusion applications, the heated melt behaves as a non-Newtonian pseudoplastic material. Its viscosity decreases with increasing shear rate and temperature, and also with increasing concentrations of water and other viscosity-reducing agents.

[0020] Food extrusion can be divided into two general categories: forming and cooking. In forming applications, low shear extruders are primarily used to mix and form the desired product shape with minimal energy input. Product applications include pasta, cold-formed snacks, and other raw cooked pellets. The screws typically have deep channels (0.2-0.3D) with no compression ratio and operate at relatively low speeds (below 50 rpm).

[0021] Cooking extrusion applications typically use medium- and high-shear single- and twin-screw co-rotating extruders. Viscous heat dissipation, heat transfer through the barrel, and sometimes steam injection can provide significant energy to the product. The screws operate at relatively high speeds (>100 rpm), the channel depth is shallow, and single-screw extruders have compression ratios of up to 4:1. These extruders are used for cooking and shaping expanded products.

[0022] Both single-screw and twin-screw extruders are used in the extrusion of plastics and food products. Twin-screw extruders offer significant advantages over single-screw extruders in feeding, mixing, heat transfer, residence time distribution, and pump performance. However, twin-screw extruders are more expensive. The primary applications of twin-screw extruders in polymer extrusion are: (i) profile extrusion of heat-sensitive materials, such as polyvinyl chloride (PVC); and (ii) compounding and difficult polymer processing operations requiring good mixing, volatilization, and chemical reaction of various compounds.

[0023] Single Screw Extruder

[0024] Figure 4a / b shows an example of (a) a thermoplastic single screw extruder, (b) a thermoplastic single screw extruder. Figure 5 provides an explanation of screw terminology using the screw elements of a single screw extruder.

[0025] Twin Screw Extruder

[0026] Regarding twin-screw extruders, prior art discussions have been limited to intermeshing twin-screw extruders. There are two types: co-rotating and counter-rotating, depending on how one screw rotates relative to the other. One major difference between single-screw and twin-screw extruders is the type of transport that occurs within the extruder. Single-screw extruders rely entirely on frictional resistance flow in the solids transport zone and viscous resistance flow in the melt transport zone. Twin-screw extruder transport is less dependent on the frictional properties of the material due to the action of the second screw in the intermeshing zone, which provides some positive displacement. Intermeshing counter-rotating twin-screw extruders provide the most reliable displacement action.

[0027] The channels of a co-rotating twin-screw extruder cannot physically close completely in the intermeshing region (see Figures 6-8). This is determined by the nature of the screw rotation. Due to surface wiping of the screw surfaces in the intermeshing region, material is wiped off one screw surface, changes direction, and moves to the other screw (see Figure 9). Because the channels are open axially, pressure differences within the extruder can act on the material and cause it to move backward.

[0028] In fully intermeshing counter-rotating extruders, the screw channels are closed both radially and axially. The flight of the second screw in the intermeshing region physically prevents material from moving to the other screw. Each revolution of the screw thus creates a C-shaped closed chamber (along each screw) that moves one pitch (see Figures 6 and 9). Thus, these extruders are not vulnerable to pressure flow and can generate higher die pressures than other extruders under similar conditions. Material mixing is limited to the flow profile within the C-shaped closed chamber and leakage flow between the screw and barrel components.

[0029] Material flow in a co-rotating twin-screw extruder follows the profile shown in Figure 8, with corresponding changes in direction and surface renewal. This results in a relatively uniform shear stress distribution around the screw (see Figure 10). The screw configuration typically includes mixing and flow-restricting elements. This creates complex flow patterns, resulting in good mixing and heat transfer, large melt capacity, and good melt temperature control. The complexity of the flow patterns makes modeling and scale-up more difficult. Prior art scale-up has been achieved in part through modular extruder designs, where the screw configuration can be easily adjusted to achieve desired performance characteristics.

[0030] The shear stress distribution around the screws of a counter-rotating extruder is not uniform (see Figure 10). Trapped material in the intermeshing area can push the two screws outward toward the barrel, causing excessive wear. This limits melt extrusion operation at relatively low viscosities and low screw speeds. Its strength coincides with high volumetric transfer. Twin-screw extruders are generally starved, with the feed system controlling their output. The degree of screw filling depends on the flow rate, material viscosity, screw design, screw speed, and the pressure profile generated by the limiting screw element or die assembly. Therefore, the mechanical energy input depends on the extruder design, shear rate, and material viscosity. It should be noted that the material flow profile in a co-rotating twin-screw extruder is complex and technically much more difficult to capture through modeling or simulation than that of a single-screw extruder.

[0031] In the prior art, twin-screw extruders were first used in food preparation applications in the 1970s. Since then, their use has increased at the expense of single-screw extruders and the expansion of general food extrusion applications. As previously mentioned, co-rotating twin-screw extruders have greater volumetric displacement and mixing capabilities than single-screw extruders. This has allowed for the development of improved process capabilities and flexibility with more consistent food quality. They have also been adopted for low- and medium-viscosity, as well as high-viscosity applications. Modular extruder design and superior process control have expanded their use to applications requiring multiple process and functional zones. These include multiple liquid injections, downstream solids feeds, and vent ports. They can be designed with relatively long barrels to provide the long residence times required for specific conversions and reactions. The flexibility of extruder design has also accelerated the development and optimization of new, more challenging products, even when the exact flow behavior is not fully understood. Figure 11 shows an example of the potential capabilities of twin-screw extruders for multi-process extrusion systems. It shows two dry mix feeders, liquid and direct vapor injection ports, a side feeder for other solid ingredients, a vent port with vent stuffer, an injection port for heat-sensitive liquid ingredients, a start / throttle valve, a die plate, and a die face cutter assembly. In reality, most extrusion systems are much simpler and may not include all of the units shown in Figure 11.

[0032] Counter-rotating extruders have been used for low-viscosity applications requiring high-pressure molding capabilities, such as candy and licorice. They typically rely to a large extent on barrel heat transfer to heat and cook the material. The trapped material in the C-shaped chambers can be heated to high temperatures without steam blowing back into the feed port. Pressures in these chambers can be raised and maintained to levels higher than the corresponding local steam saturation pressure.

[0033] Considerable efforts have been made in the prior art to provide modeling and simulation structures that capture co-rotating twin-screw extruders. Most of these have emphasized establishing empirical relationships correlating the effects of feed ingredients and operating conditions on product properties. Twin-screw designs typically incorporate mixing disks, paddles, or counterscrew elements. As food compositions become more complex, the task of detailed engineering modeling becomes even more challenging. The prior art offers very few modeling approaches. For example, while there is prior art simulation modeling of twin-screw co-rotating extruders, the modeling structure is unidirectional and only considers non-Newtonian and non-isothermal melt rheology.

[0034] molding

[0035] The forming process is understood as the process of forcing the melt through a specially designed die orifice. Depending on the die geometry, melt viscosity, and flow rate, a constant pressure drop ΔP ≒ Qμ / K occurs across the die, where K is the die conductance coefficient, equal to the inverse of the die resistance. However, this relationship typically assumes laminar flow through the die without material slippage on the metal surface. Slippage results in a plug-flow-like pattern and can be promoted by using smooth, low-friction die materials. Heating the die can further reduce the friction coefficient and therefore further increase slippage. Material slippage is expected to result in a lower pressure drop across the die, which can affect total extruder fill, mixing intensity, and energy input. From a product formation perspective, the primary benefit of die slippage is the creation of a smooth, undamaged product surface. This is why Teflon® inserts are often used in dies for pasta products.

[0036] Analysis of food melt rheology at different shear rates (volumetric output) is more complex than that of thermoplastic materials. Changes in melt properties due to changes in feed rate affect the time-temperature history of the extruder and must be taken into account.

[0037] In most food extrusion operations, the extrudate swells after leaving the die, as shown in Figure 12. The large and complex molecular structure of polymers results in elastic properties, which is believed in the prior art to be the main cause of this phenomenon. Swelling is a form of elastic recovery from the deformations that the polymer undergoes in the die. Swelling increases with sudden die flow inlet and shortening of the die land length. It also increases with increasing shear rate in the die. In the prior art, for a circular die, the shear rate is usually calculated as γ=Q / πr 3 is modeled by an approximation of

[0038] Prior art die design and development for complex extruded food shapes was based on logical and practical methods rather than analytical approaches. These shapes are further developed for cold extrusion of pasta and semi-finished products. In culinary extrusion applications, streamlining the material flow is important to avoid material stagnation, which can cause poor quality, process instability, and potential blockage of small die areas.

[0039] mixture

[0040] The fundamental mechanism of mixing in polymer extruders is convective motion in the laminar flow region. Mixing generally occurs through shear and extensional flows. When the mixing components do not exhibit a yield point, it is called distributive mixing. Distributive mixing can be described by the degree of deformation or strain experienced by the fluid elements. Dispersive mixing, on the other hand, resembles a milling action and addresses the breakdown of solid particles and agglomerates. In this type of mixing, the actual stress is important in order to overcome the yield stress of the solid components. Distributive mixing is always present when dispersive mixing occurs, but the reverse is not necessarily true.

[0041] Distributive mixing in extruders is commonly simulated by determining the velocity profile occurring within the screw channel. In most simulation techniques, the fluid is considered Newtonian, the components have the same flow profile, and flow through flight clearances is ignored. Mixing performance improves with increasing pressure-to-counterforce flow ratio (throttling ratio). In choke-fed extruders, this reduces extruder output. In all cases, the mean residence time increases and the residence time distribution broadens (increased backmixing).

[0042] In single-screw extruders, special mixing elements are designed to promote distributive mixing. Types of distributive mixing elements include pin, Dulmage, Saxton, pineapple, slotted flight, and cavity transfer mixing sections. Quantitative analysis of these mixing elements is very difficult, and the development of these devices has been largely empirical using experimental evaluation methods. When used in thermoplastic single-screw extruders, they are typically located at the discharge end of the metering section. Modular single-screw and twin-screw extruders use kneading paddles, cut-flight screws, and discs to promote distributive mixing. These elements can be placed in any logical location within the extruder.

[0043] Dispersive mixing is known to be important in compounding and some food extrusion applications. The breaking stress of an agglomerate depends on its size, shape, and nature. The stress acting on the agglomerate depends on the flow and rheological properties of the polymer field. The higher the viscosity, the greater the dispersive mixing. Also, high-speed solid-solid mixing can cause a higher level of agglomerate breakage than liquid-solid mixing. Several dispersive screw elements are used in extruders. These include Union Carbide, Egan, Dray, and Blister Ring. In all of these cases, the agglomerates must pass through a specially designed clearance between the mixing element and the barrel. They are subjected to high shear stresses, which break them down and promote mixing.

[0044] Twin screw extruders can be equipped with wide kneading paddles to promote dispersive mixing. It is not possible to create a tight clearance dam area in an intermeshing twin screw extruder with side-by-side disks of the same diameter. In this case, the maximum disk diameter is equal to the center distance between the two screw shafts. The created tolerance / gap with the barrel is equal to half the depth of the screw channel. A tight clearance dam can best be created by placing two pairs of intermeshing disks (one behind the other) on the screw shaft (see Figure 14).

[0045] Modeling pressure development in extruders

[0046] As mentioned above, most single-screw and twin-screw extruders for food applications are typically underfed. Their throughput is determined by the feed system. Therefore, the screws in the solid-conveying section remain partially filled with negligible mechanical work done on the feed material. Screw filling occurs in successive sections due to (i) a decrease in the conveying capacity of the feed screw or kneading elements or (ii) the placement of obstacles in the form of kneading elements, disks, counterscrews, or die plates. As resistance to flow increases, the feed materials (solid and liquid) are compressed, mixed, and transformed into a fluid / melt through a hydration process involving solid-solid / solid-fluid friction and dissipation of mechanical energy. This transformation / melting process can involve a significant temperature increase, depending on the friction field and the viscosity of the material.

[0047] The length of the conveying screw section to generate the pressure necessary to overcome the restriction is affected by the screw design, the effective material viscosity, and the screw speed. The pressure gradient for a Newtonian fluid is:

number

[0048] For Newtonian fluids, the pressure gradient across the resistance section created by the kneading elements, disks, reverse screws, or die plates can again be modeled by the relationship ΔP ≈ Qμ / K, where Q is the melt volumetric flow rate and K is the conductance coefficient of the restriction section. For the reverse kneading / screw element section, K is affected by the created helix angle, channel depth, screw gap, and screw diameter. At steady state, this section is always full, and the pressure drop within the section causes a forward flow of material. The pressure flow must be greater than the resistance flow component of the restriction element in this section.

[0049] For discharge die obstructions such as circular orifices, K is, for example, K=πr 4 / 8L, where r is the die radius and L is the die land length. The pressure drop in the resistance section increases with increasing flow rate, viscosity, and section restriction. Figure 15 shows an exemplary twin-screw extruder showing simulated pressures and temperatures within the extruder.

[0050] Modeling energy consumption and supply in extruders.

[0051] Modeling energy consumption and supply is complex due to the various processes and sources within the extruder. The following must be considered:

[0052] (a) Energy supply in the extruder: The energy supplied in the extruder is mostly consumed by the product. There are also unavoidable losses from the barrel, gearbox, and motor. The energy consumed by the product is usually simulated by modeling the following changes: (1) Enthalpy rise q h (e.g., in kW). In cooking extrusion applications, the increase in enthalpy accounts for the largest energy consumption; (2) the heat of reaction / transformation, such as the heat of starch gelatinization and molecular breakdown. Gelatinization energy has been measured for a variety of starches and is generally estimated to be in the range of 10-20 kJ / kg. These values ​​are usually measured at excess moisture concentrations, which only consider the gelatinization endotherm. In high-viscosity extrusion applications (low moisture concentrations), starch and other biopolymers undergo a fractionation (depolymerization) process that is expected to consume some energy. This energy is difficult to measure and therefore difficult to quantify. (3) the potential energy q P (e.g. in kW).

[0053] The kinetic energy of the material is usually assumed to be too small and therefore ignored. Energy losses are: (1) from the barrel by the circulating liquid in the barrel jacket, q cl (1) from the barrel in the form of convection cooling by ambient air; (2) from the gearbox and motor. This occurs in the form of forced cooling to remove the heat generated and maintain the gearbox and motor at an acceptable temperature. Losses can be estimated by calculating the applied cooling load. They typically range from 5%-15% of the motor energy, depending on the size and operation of the extruder.

[0054] (b) Energy Supply: Energy is supplied to the material from the extruder motor, barrel heat transfer, and in some cases through direct steam injection.

[0055] (c) Mechanical Energy Input: Mechanical energy is supplied by the extruder motor. It is mostly consumed by the material in the filling section of the extruder through friction and viscous heat dissipation. For Newtonian fluids, the mechanical energy input (E) is

number

[0056] It can be observed that viscous heat dissipation can be increased by increasing the length of the screw filling section, material viscosity, and screw speed. It is noted that increasing the screw speed results in a decrease in the length of the melt pump section. On the other hand, the resistance section (disk, reverse kneading element, and screw) is always filled, and the filling length is constant, regardless of the screw speed or feed rate.

[0057] In the prior art, the total specific mechanical energy input (kWh / kg) of a direct current (DC) motor with a constant maximum torque is estimated or approximated using a modeling structure such as S = (%N·%T·p) / F, where N is the maximum extruder screw speed (rpm), T is the maximum motor torque (Nm), P is the total installed motor power (kW), and F is the feed rate (kg / h). The screw speed and torque are proportional to the motor armature voltage and current, respectively. For alternating current (AC) motors, the actual power output should be obtained from a wattmeter.

[0058] (d) Heat transfer through the barrel: The rate of heat transfer (kW) to or through the extruder barrel is a h =U A ΔT lmcan be modeled or approximated using the modeling structure: 2 °C), A is the internal barrel surface area (m 2 ), ΔT lm is the logarithmic temperature difference (°C) between the inner barrel surface and the material in contact with it.

[0059] The heat transfer coefficient in the unpacked section is relatively small due to the lack of intimate contact with the surface. The heat transfer coefficient i is affected by the material properties and the mixing pattern. It generally increases with decreasing viscosity and increasing screw speed. Levine and Miller (2007) presented a review of published research on barrel heat transfer. Heat transfer coefficients range from 50-300 W / m 2 It is estimated to be 100°C.

[0060] The role of barrel heating becomes important in low viscosity extrusion cooking applications, such as those containing high moisture, sugar, and / or fat concentrations. In such applications, the proportion of mechanical energy input is reduced. In high melt viscosity applications, mechanical energy from the motor is typically the dominant source.

[0061] Heat can also be transferred from the material by applying cooling to the barrel jacket. Cooling the product increases the melt viscosity and increases the rate of mechanical energy input. Therefore, the net product cooling is lower than the overall cooling load removed from the barrel.

[0062] Barrel heat transfer per unit of output can be greater in small extruders than in large extruders. In either case, barrel temperature management is important, even if the heat transfer rate is small. As mentioned previously, the coefficient of friction of a material with a metal surface can be affected by the surface temperature.

[0063] (e) Steam Injection into the Barrel: Water is usually a natural component of the overall food formulation. A portion of the total water can be injected directly into the extruder in the form of steam. This can provide a significant amount of energy, thus supplementing and / or replacing a portion of the motor and barrel heating energy. In the prior art, for example, steam energy q s (kW) modeling is

number

[0064] Modeling the effects of screw / barrel wear on extruder performance

[0065] Extruders function as both pumps and thermomechanical reactors. These functions are influenced by the screw design, including the various gaps within the extruder. With use, screw elements and barrel materials undergo a process of wear. The wear rate depends on abrasives, adhesives, and corrosives; component materials of construction; and operating conditions. Increasing the gap size in the forward conveying section reduces the extruder's pumping efficiency and increases the extent of backmixing and retention time. This can increase the total mechanical energy input to the product. This can also reduce extruder output when operating in or near choke-feed mode.

[0066] Additionally, restrictive sections within the extruder and die assembly can also wear. The resulting larger gaps reduce the flow resistance of these sections, requiring a lower pressure drop across them. This can reduce the extruder charge length of the advancing pump section and the total energy input to the product.

[0067] For heat-sensitive materials such as food, larger gaps and possible material blockages can potentially result in product degradation and poor quality, possible die clogging, and process instability.

[0068] Adjustments to product recipes (especially water addition) and operating conditions can, in some cases, compensate for gradual wear and extend the useful life of extruder components. Screw profiles may also be modified to adjust the length of the melt-fill section. In scenarios where the extruder length is longer than necessary, the fill (working) section can be moved up or down the screw to use unworn sections of the barrel liner. However, this requires proper process knowledge of the effects of screw configuration on process and product performance. Eventually, a situation will be reached where product and / or output specifications cannot be met, necessitating part replacement. New parts may require re-adjustment of extruder conditions if changed during the part's wear life.

[0069] Ventilation inside the extruder

[0070] Venting in cooker extruders can be very useful for several purposes, such as accelerating cooling, removing certain volatile components, and influencing certain changes in the product. To allow for venting in the non-filled sections, specially designed screws are required. Typically, a high volumetric flow capacity screw section is placed with a vent port following a restrictive / low volumetric flow capacity screw section that acts as an airtight seal (see Figure 16).

[0071] Melt rheology and temperature affect the behavior of the melt and the degree of removal of volatile components from the vent section. Melts with relatively low viscosity (e.g., those with high initial moisture content) can be aerated successfully. If the initial moisture / fat content is reduced and / or excessive steam is released, the aerated melt may create a highly expanded mass that is difficult to convey forward. In certain situations, the use of a stuffer (see Figure 17) may be helpful.

[0072] In applications such as cooked semi-finished products, significant product cooling is required to avoid product expansion and achieve good die face cut. The optimum extrusion process design depends on the nature of the formulation and the degree of cooking desired. If this can be achieved at relatively low temperatures, cooking, cooling, and shaping can be performed in a single extruder, with or without the use of vents.

[0073] For relatively high temperature (e.g., above 135°C in the cooking section) and high capacity applications, the use of a "two extruder" system is usually preferred. The first extruder cooks the material. Aeration and cooling occur between the two extruders. The second extruder compresses the expanded material and forms the cooled product into the desired shape. In this way, the two extruders can be designed and operated independently. In this way, the cooking and forming extruders are utilized for maximum throughput and efficiency without compromises, as would be the case with a single cooking / forming extrusion system. The forming extruder is usually a single screw type similar to those used in pasta product applications.

[0074] Conclusions on the extruder simulation and modeling process and its usefulness in extrusion automation and autopilot operation of industrial extruders

[0075] Extruders are common devices in the plastics, metal, and food processing industries, and the use of the extrusion process is particularly widespread in product manufacturing and food processing. Extruders are particularly popular in food processing because of the creation of a fast, continuous process that can be used in the food industry to produce many foods, such as snacks, breakfast cereals, pelleted products, pet food, and pregelatinized flour. It is a system that encompasses multiple unit operations, such as mixing, kneading, cooking, forming, and cutting, all in one device.

[0076] Generally, primary extruders can be classified as single-screw or twin-screw types, with the former being more widely applied in general polymer processing and the latter being more widely applied in the compounding of various fibers, fillers, and polymer blends prior to final molding. Single-screw extruders often consist of only one screw housed in a barrel with a grooved or fluted design. Furthermore, the screws in single-screw extruders are typically designed to have a reduced pitch to create compression. The amount by which the pitch is reduced is called the compression ratio.

[0077] Twin-screw extruders can be further classified into two types based on the interaction of the two screws: intermeshing and non-intermeshing. Within the twin-screw extruder family, fully intermeshing counter-rotating twin-screw extruders have been found to have the best pumping capacity due to their volumetric displacement characteristics. Therefore, twin-screw extruder screw sets can be either co-rotating or counter-rotating. Co-rotating extruders are more frequently used because they can impart more mechanical energy to the material than counter-rotating screws. Twin-screw extruders require higher maintenance costs than single-screw extruders, but are more commonly used in the food industry due to their wide range of operating conditions and ability to produce a wide range of foods. Twin-screw extruders have greater flexibility for handling a variety of ingredients and higher production rates than single-screw extruders. Twin-screw extruders can operate over a wide range of moisture contents, a drawback of single-screw extruders. Preconditioning systems can be used to expand the capabilities of both single-screw and twin-screw extruders. Twin screw extruders have high mixing efficiency, self-wiping capabilities to prevent residue buildup, and relatively faster and more uniform heat transfer from the barrel to the ingredients.

[0078] In principle, extrusion is a multiple-input, multiple-output (MIMO) system. Figure 3 illustrates the inputs and outputs relevant to this concept. Various extrusion process parameters can be broadly classified into three categories: (1) independent parameters (input parameters), (2) system parameters (dependent parameters), and (3) product characteristics (output parameters). Independent parameters are parameters that the extruder operator can directly control. These include raw material (feedstock) characteristics and extrusion operating parameters such as feed rate, barrel temperature, screw configuration, screw speed, and die dimensions. By modifying the independent parameters, the operator can achieve changes in system parameters and final product characteristics. System parameters are parameters that the extruder operator cannot directly control, but the operator can influence them by modifying the independent parameters. Mean residence time, residence time distribution, backpressure, motor torque, and the specific mechanical energy applied to the material are measurable system parameters. These affect final product characteristics but can only be changed indirectly by modifying the independent parameters. Product characteristics are parameters that help describe the quality of the final extruded product. This can include physical properties (expansion ratio, density, etc.), chemical properties (water absorption and solubility, etc.) and sensory properties (crispness, crispness, texture, etc.) In order to steer, optimize and / or control the extrusion process, e.g., for consistent production of food products and / or optimization of energy consumption, the MIMO system needs to be captured, parameterized and controlled.

[0079] In modeling food and polymer extrusion processes, as mentioned above, models are generally deterministic approximations based on transport phenomena of distributed parameters or locally lumped parameters. In engineering, a lumped parameter approach is commonly used. In the prior art, the primary objective of extrusion engineering design is to predict the pressure and average process material temperature along the machine as a function of operating conditions for a given screw / die geometry. In these model structures, the screw channel is divided into short segments, the input data comes from the calculation of the previous segment, and the output data from the current segment is the input data for the next segment. Within a segment, local parameters are assumed to be constant. This locally lumped parameter concept is particularly used when dealing with plasticization processes such as extrusion, where solid transport and melting are modeled in addition to melt flow.

[0080] Computer modeling is commonly used to answer technical questions in food and polymer extrusion. It is used to study the effect of input parameter changes on material / process output and to plot trends in material response to hundreds of potential input data configurations. It is used to obtain process maps for identifying required process parameters according to the target material and selected screw profile: distribution of thermomechanical results (temperature, pressure, viscosity, and packing) along the screw, detailed energy balance of the process, residence time analysis, etc. Another technical application is the software's scale-up module, which converts processes to larger or smaller diameter machines (process scaling). Starting from a model process, the target design geometry and related process parameters are technically simulated / modeled based on conversion rules. However, practical use suffers from many drawbacks. Importantly, the global modeling structure describing the complete extrusion process, including solid transport, polymer melting, and melt flow, is based on relatively simple 1D or 2D models, which do not allow for detailed exploration or technical capture of the process. Furthermore, the physics within extrusion is not always fully understood and / or is often too complex to be captured and / or simulated by implementable modeling structures. However, even when the physics is known and can be modeled, modeling approaches often technically fail due to the complexity and uncertainty of multi-parameter processes, and therefore require low accuracy or enormous processing power to be able to operate within an acceptable timeframe.

[0081] In the prior art, Patent Document 1 discloses a system for controlling the thickness of a sheet produced by extruding a material through a die having multiple thickness adjusting means by repeatedly applying the following steps: 1) measuring the thickness distribution in the transverse direction of the sheet; 2) predicting future changes in the sheet thickness using an evaluation function based on a process model representing the relationship between the manipulated variables, the sheet thickness, and the sheet thickness measurements, and deriving a manipulated variable time series so that the evaluation function is minimized; and 3) outputting at least the initial manipulated variables of the derived manipulated variable time series to a thickness adjusting means. Patent Document 2 discloses a system for producing an elastomer composition, including a step of injecting a raw material into an extruder, wherein the injection step is adjusted based on: i) an actual weight value of the raw material measured at an injection instant preceding a given injection instant; ii) an expected weight value of the raw material calculated for the corresponding injection instant preceding the injection instant; and iii) an expected weight value of the raw material calculated for a predetermined prediction period following the injection instant. This system attempts to minimize the predicted weight error between the actual weight measured during the prediction period and the expected weight calculated for the prediction period, and the model weight error between the actual weight measured during the prediction period and the theoretical weight of the ingredient corresponding to the target weight loss of the ingredient. Non-Patent Document 1 provides a review of advances in food processing using ANNs, including application areas of ANNs ranging from shallow learning to deep learning. In particular, the state-of-the-art technologies of machine learning, deep learning, and image processing for food processing are discussed. Furthermore, Patent Document 3 discloses a manufacturing system including: (i) generating candidate manufacturing conditions for producing a product; (ii) using a predictive model to determine a prediction of the production results when the product is produced under each of one or more candidate manufacturing conditions; and (iii) generating an evaluation of each of the one or more candidate manufacturing conditions by evaluating the results of the prediction based on predetermined evaluation criteria. This process is repeated while switching between the candidate manufacturing conditions, and (iii) determining the candidate manufacturing conditions that meet the predetermined criteria as the manufacturing conditions for producing the product. Patent Document 4 discloses a resin film manufacturing system equipped with lip gap control that measures the difference between the state of a target heat bolt (THB) at the start of current control and the state of each of the heat bolts at the start of final control.If the difference between the state of the THB at the start of the current flow and the start of the final control is minimal, the learned result of the THB is set as the initial value of the control condition of the THB. If the difference in the state of the THB is not minimal, either the learned result of the heat bolt or the learned result of the THB, whose state difference is smaller than the difference in the state of the THB, is set as the initial value of the control condition of the THB. Patent Document 5 discloses an extruder for extruding food products. A starting product and water are supplied to the extruder and mixed in a mixing zone. The mixture is heated and plasticized by the supply of shear energy, and foamed after passing through a nozzle. To achieve a constant product quality that is independent of changes in the characteristics of the starting product and external variables, the drive force applied by the drive motor of the shaft or shafts at a constant shaft speed of the extruder is measured, the mass throughput of the supplied starting product is measured, the quotient of these variables is measured, the mass throughput of the supplied starting product is measured, the quotient of the drive force applied by the drive motor of the shaft or shafts is measured, and the mass throughput of the supplied starting product is measured. The quotient of these variables is formed, and the specific energy determined in this way is used as a control variable. The energy thus determined is provided as a control variable to a control device, which maintains it constant according to a predetermined set point, such that the water or steam supply to the mixing zone of the extruder is affected as a control variable. Finally, Patent Document 6 discloses an injection molding system including a barrel connected to a hopper for receiving material from the hopper. The system includes a heater outside the barrel and an extrusion screw inside the barrel. The system includes a motor coupled to one end of the extrusion screw for rotating the extrusion screw, a torque sensor on the motor, and a controller coupled to the motor and the heater. The controller receives a signal from the torque sensor. The controller includes a control algorithm that adjusts the heater according to the signal from the torque sensor to melt the material in the barrel. [Prior art documents] [Patent documents]

[0082] [Patent Document 1] EP1319492A1 [Patent Document 2] US2010 / 0149902A1 [Patent Document 3] US2020 / 0293011A1 [Patent Document 4] US2022 / 0072756A1 [Patent Document 5] EP0265601A2 [Patent Document 6] US2016 / 0158985A1 [Non-patent literature]

[0083] [Non-Patent Document 1] J. Nayak “Intelligent food processing: Journey from artificial neural network to deep learning” Computer Science Review, 2020 Summary of the Invention

[0084] The objective of the present invention is to provide an industrialized system and method for the automation and / or optimal control of industrial extrusion processes. The system should be capable of optimizing food and / or polymer processing through extrusion techniques directed at all of the multiple unit operations, including mixing, kneading, shearing, cooking, molding, and forming. It should be noted that the extrusion cooking process is typically a complex, high-temperature, short-duration process in which food materials are cooked in a tube through a combination of moisture, pressure, temperature, and mechanical shear, resulting in molecular transformation, gelatinization, protein denaturation, and bond breaking, leading to products with novel shapes and textures. The system should also be capable of optimizing extrusion by leading to lower energy consumption, reduced antinutritional factors, improved product microbiological safety, and ultimately improved consumer acceptance. Furthermore, the system should be capable of automatically adapting and optimizing raw material characteristics, including thermophysical properties and flow properties, under a range of conditions within the extruder, which are essential for controlling extruder behavior for a high-quality end product. The system should also be capable of considering thermophysical properties such as material density, specific heat, thermal diffusivity, and thermal conductivity, which provide a basis for modeling for the design and optimization of food processing operations involving heat and mass transfer. The system should also be able to handle input measured parameters as physicochemical properties, which are often important for capturing, characterizing, or measuring changes during extrusion in processed materials, such as food materials, and can account for thermal and mechanical effects and modifications thereof. Finally, the system of the present invention should be able to capture and dynamically respond to the effects of extrusion variables (feed moisture content, screw speed, and barrel temperature) on the proximate, thermophysical, physicochemical, extrudate, color, texture, and sensory properties of processed materials, such as food and / or polymeric materials.

[0085] According to the invention, these objects are achieved in particular by means of the features of the independent claims. In addition, further advantageous embodiments can be derived from the dependent claims and the associated description.

[0086] A commonly known extruder system comprises a feeder, an extruder, a forming orifice (die), a collecting means, and an extruder control, wherein the feeder supplies a plastically deformable and / or viscous input material to the extruder, the extruder continuously forcing the input material into and out of the forming orifice forming an output material as extrudate, and the collecting means collects the extrudate for further processing, the extrusion process being controllably steered during operation by the extruder control, which includes programmable logic for setting and adapting operational setting parameter values ​​(208) of the operating units (105) of the feeder (101), extruder (102), forming orifice (die), and collecting means.

[0087] This known extruder system is controlled manually or semi-automatically, and in particular, the process parameters for controlling the system are set by the extruder control unit based on the experience and expertise of the extruder system operator. On the one hand, increasing demands on the material property parameters / properties of the extrudate, and on the other hand, energy-efficient operation of the extruder system, mean that such extruder control, operated manually or semi-automatically by an operator, cannot ensure optimal control of the extrusion process and material property parameters. For example, plant-based protein extrudates have very high requirements and strict tolerances for material property parameters because these products are directly compared to meat / fish products. On the other hand, the operating control parameters have a nonlinear relationship with the input parameters used to control the extrusion process. Furthermore, knowledge and experience of the extruder system require the presence of an operator, and in particular, a change of operator causes at least partial loss of the operator's knowledge and / or experience.

[0088] According to the invention, this object is achieved in particular by the elements of the independent claims. Further advantageous embodiments are evident from the dependent claims and the description.

[0089] In particular, these objects are achieved by an extruder system having a feeder, an extruder, a forming orifice (die), a collecting means, and an extruder control, wherein the feeder supplies a plastically deformable and / or viscous input material to the extruder, which extruder continuously presses the input material into and out of the forming orifice forming an output material as extrudate, and the collecting means collects the extrudate for further processing, the extrusion process being controllably steered during operation by the extruder control, the extruder control having programmable logic for setting and adapting operational setting parameter values ​​of the operating units (105) of the feeder, extruder, forming orifice (die), and collecting means, the extruder control further having a digital controller for signaling and steering the programmable logic, wherein to operate the programmable logic, the digital controller takes in input parameter values ​​including at least process parameter values ​​and / or operational setting parameter values ​​and / or material property parameter values ​​and / or environmental measurement parameter values, the input parameter values ​​being controlled by the extruder and / or extruder and / or or sensory parameter values ​​measured by sensors associated with the forming opening and / or the collecting means, the extruder control has a repository storage unit (with an adaptive digital database (504)) holding a plurality of selectable structured data records for storing digital recipes, each of the selectable data records including at least material property parameters of the input material and target material property parameters of the extrudate (300) and / or initial operational setting parameters providing an initial setting of operational setting parameters (201) for operation (105) of the extruder system (1), the input parameter values ​​(200) further including the parameter values ​​of the selected record (507), and the extruder system (1) includes digital signals for steering the programmable logic (501) and associated operational units (507) by the digital controller (502) to controllably and steerably extrude an extrudate (300) having material property parameter values ​​(202) within a predefined tolerance range of the predefined target parameter values ​​(206).

[0090] The digital controller includes a database of recipes for extruding extrusion data having predefined material properties, and the recipes are adapted to set operational setting parameter values ​​to match the extruder controller to ensure production of the predefined extrusion data. Preferably, the recipes include initial operational control parameters for starting the extruder system to ensure optimal process parameters and thereby ensure a stable and energy-efficient extrusion process. Furthermore, the input parameters include environmental measurement parameters for characterizing the environment of the extruder system, such as temperature, humidity, and weather forecast data for the extruder system environment. The environmental measurement parameters are taken into account in optimizing the energy consumption of the extruder system.

[0091] The adaptive digital database of the extruder system is realized as a digital library, and the repository storage unit has a network interface that provides access to the structured data records via a data transmission network for selection and / or adaptation and / or generation of structured data records to ensure data exchange and data tracking of data sent, received and generated by the digital controller and repository storage. Structuring the data records allows searching the digital database and filtering the data records according to structural criteria.

[0092] The digital controller includes a machine learning unit that monitors and classifies input parameter value patterns and adapts operational setting parameter values ​​of the operational units of the feeder and / or extruder and / or forming orifice (die) and / or collecting means to align the measured material characteristic parameter values ​​of the extrudate within predefined tolerance ranges, enabling the digital controller to build a model of the extruder system based on sample data, known as training data, to make predictions based on the input parameter values ​​of the operational setting parameters for extruding extrudates having material characteristic parameter values ​​within predefined tolerance ranges of predefined target parameter values, without being constrained by a linear relationship between the input parameters and the operational setting parameters.

[0093] The machine learning unit has at least a deep learning (DL) structure including one or more neural network (NN) structures and / or one or more statistical modeling structures, and provides output parameter values ​​based on input parameter values ​​that indicate the parameter value adaptations required to adjust the measured material property parameter values ​​of the extrudate within a predetermined tolerance range to ensure that the machine learning model is a nonlinear function, a piecewise linear function, or a step function.

[0094] Machine learning-based deep learning architectures comprise at least a cascade of multiple layers of nonlinear processing units for feature extraction and signal transformation, with each successive layer using the output of the previous layer as input, providing supervised learning for classification and / or unsupervised learning for pattern recognition, ensuring accurate modeling of complex nonlinear models. Increased computational power allows for efficient training of several dependent layers. Supervised learning is preferred when input data is collected or generated by simulation by example. Another preferred option is to apply unsupervised learning models, which, in contrast, operate on their own to discover the inherent structure of unlabeled input parameters.

[0095] The extrusion process of the extruder system is adapted autonomously by the machine learning unit by automatically adapting the operating setting parameter values ​​of the feeder and / or extruder and / or forming opening and / or collecting means operating units, and by time-based monitoring of input measured parameter values, adjusts the measured material property parameter values ​​of the extrudate within predefined tolerance ranges, ensuring that the extrusion process is controllably steered in the most effective way by the digital controller taking into account the input parameter values.

[0096] Faults in the extruder system during the extrusion process are automatically detected by the machine learning unit based on the measured and monitored input parameter values, and warning and / or steering signals are generated upon detection of predicted faults in the extrusion process, ensuring a controlled extrusion process that produces extrudates according to predefined material property parameters. Furthermore, due to the intermediate interaction of the digital controller with the warning and / or steering signals, there is the most effective use of input material for extrusion of extrudates by the extrusion system. Furthermore, there is inefficient consumption of energy due to the detection of predicted faults in the extrusion process.

[0097] Deep learning for structural extrusion systems involves, at a minimum, a convolutional neural network (CNN) structure as a deep neural network that ensures local spatial consistency within input parameters, allowing some parameters to have fewer weights since they are shared. This process, in the form of convolution, is particularly well-suited for extracting extrusion process-related information at low computational cost of input parameters, especially electromagnetic-based data, including optical data generated by, for example, a near-infrared sensing unit.

[0098] The measured and / or captured input parameter value patterns are classified and selected by the convolutional layer and pooling layer of the CNN structure, and the pooling layer reduces the dimension of the feature map of the measured and / or captured input parameter value patterns, thus reducing the processing complexity within the machine learning unit to adapt the operational setting parameter values ​​of the operational unit.

[0099] The repository storage unit of the extrusion system further includes an operational data storage for storing historical operational data including historical input parameter values, historical material property parameter values, and predefined target parameter values, and the machine learning unit is trained by applying the historical operational data to ensure that supervised training using the historical operational data for most effective modeling of the deep learning structure is most effectively adapted to the past behavior of the extrusion system characterized by the historical operational data. Furthermore, a centralized database communicating / exchanging data with multiple extruder systems can be accessed (read) by at least one extruder system via a data transmission network, increasing the data used to train the centralized deep learning structure that can be applied by at least one extruder system.

[0100] The material property parameters include texture and / or density and / or color and / or anisotropy and / or chemical composition and / or thickness and / or degree of polymerization and / or moisture content and / or protein content and / or starch content and / or fiber content and / or particle size and / or surface structure and / or tolerance ranges to ensure the most suitable properties of the extrudate produced by the extrusion system.

[0101] The operational setting parameters of the extrusion system include the screw speed of the extruder screw that presses the input material, and / or the addition rate by the feeder of at least one ingredient to constitute the input material, and / or the conditioning setting of the extruder conditioner and / or the forming orifice conditioner for cooling or heating the input material in the extruder, and / or the positioning size of the forming orifice area, ensuring effective generation of the operational setting parameter values ​​by the digital controller.

[0102] The target parameters of the extrusion system include at least one of the material property parameters and / or process parameters including at least the material property parameters and / or energy consumption of the extrusion system that ensures the most effective and sustainable extrusion process for producing the extrudate specified by the process parameters and energy consumption.

[0103] Material parameter values ​​of the input material and / or components of the input material are automatically determined by a machine learning unit which adapts the dosing process of the feeder by adapting the operational setting parameters of the feeder.

[0104] An extruder network includes two or more extruder systems and a central digital controller including a central repository with at least one adaptive central digital database containing structured data records for storing digital recipes and / or ingredients and / or products, wherein at least one of the digital controllers of the multiple extruder systems is given read and / or write access to the structured central data records via a data transmission network to select, adapt, and / or generate the structured central data records, the central data records generated by the extruder systems containing at least one data classification parameter characterizing the generating extruder system or factors affecting the generating extruder system, ensuring a centralized structured data collection of recipes and / or ingredients and / or products accessible by one or more extruder systems via the data transmission network. Furthermore, the central repository makes available know-how generated by the extruder systems via the central structured data records. Furthermore, the centralized digital controller enables a centralized deep learning architecture trained with the structured data records of several extruder systems independent of geographic location and environmental conditions.

[0105] The data classification parameters of the extruder network include the country of operation of the generating extruder system and / or the operator of the generating extruder system and / or the generating extruder system ID (system identification) and / or extruder type for classifying the data records stored centrally in at least one central database having data identifying the generating extruder system, ensuring a deeper structure of the multi-dimensional central data records that enables more effective training of the at least one deep learning structure for generating operational setting parameters based on the applied input parameters. [Brief explanation of the drawings]

[0106] The invention will now be described in more detail with reference to the following drawings:

[0107] [Figure 1] FIG. 1 shows a diagram that schematically illustrates an exemplary simplified schematic of an extrusion process, showing the conversion of raw materials (such as flour or starch) into a final product. [Figure 2] A block diagram is shown schematically illustrating an exemplary flow chart of an extrusion production line. The extrusion process begins with the characterization and receipt of raw materials. The raw materials used are critical to the consistency of the extrudate at the end of the processing line. The raw materials then undergo mixing and / or preconditioning, which can be done using equipment such as ribbon blenders and preconditioners to ensure uniformity as they enter the extruder. However, mixing and preconditioning are optional for certain products. This is followed by extrusion, which is the primary cooking step in which the raw materials are cooked and transformed into a shaped product. Post-extrusion processing operations include cutting the extruded product to the appropriate size, drying the product to the desired moisture content, and possibly seasoning or coating the product to impart a desired flavor and taste before packaging. Along with these primary processing steps, there are convolutional neural networks (CNNs) depending on their intended use. [Figure 3] 3 shows a block diagram that schematically illustrates an exemplary extrusion system as a multiple-input and multiple-output (MIMO) system. FIG. 3 illustrates various inputs and outputs associated with the extrusion system of the present invention. [Figure 4] FIG. 1 shows a diagram that schematically illustrates an example of a digital controller with an interface to programmable logic. [Figure 5] Provides a description of the input parameters. [Figure 6] Provide digital recipe instructions. [Figure 7] 1 shows a diagram that schematically illustrates an example of a neural network. [Figure 8] FIG. 1 shows a diagram that schematically illustrates an example of a convolutional neural network with a feature extraction layer and a classification layer. [Figure 9] FIG. 1 shows a diagram that schematically illustrates an example of an extrusion network having multiple extruder systems communicating over a data transmission network. [Figure 10] FIG. 1 shows a diagram that schematically illustrates an exemplary simplified schematic of an extrusion process, showing the conversion of raw materials (such as flour or starch) into a finished product. [Figure 11] A block diagram is shown that schematically illustrates an exemplary flow chart of an extrusion production line. The extrusion process begins with the characterization and receipt of raw materials. The raw materials used are critical to the consistency of the product at the end of the processing line. The raw materials then undergo mixing and / or preconditioning, which can be done using equipment such as ribbon mixers and preconditioners to ensure uniformity as they enter the extruder. However, mixing and preconditioning are optional for certain products. This is followed by extrusion, which is the primary cooking step in which the raw materials are cooked and transformed into a formed product. Post-extrusion processing operations include cutting the extruded product to the appropriate size and drying the product to the desired moisture content, as well as seasoning or coating to impart a desired flavor and taste to the product before packaging. Along with these primary processing steps, additional steps may exist depending on the type of product being produced and its intended use. [Figure 12] A block diagram is shown that schematically illustrates an exemplary extrusion process as a multiple-input, multiple-output (MIMO) system. Figure 3 illustrates various exemplary inputs and outputs associated with the extruder system of the present invention. The various extrusion processing parameters can be broadly divided into three categories: (1) independent parameters (input parameters), (2) system parameters (dependent parameters), and (3) product characteristics (output parameters). [Figure 13] 1 shows a schematic diagram of an exemplary (a) thermoplastic single screw extruder, (b) thermoplastic single screw. [Figure 14] A description of screw terminology using single screw extruder screw elements is provided. [Figure 15]1A and 1B show schematic diagrams of exemplary co-rotating and counter-rotating intermeshing screws. [Figure 16] FIG. 1 shows a schematic diagram of an exemplary intermeshing co-rotating double-flighted screw and 45° kneading block. [Figure 17] FIG. 1 shows a schematic representation of an exemplary co-rotating screw showing open channels in the meshing region. [Figure 18] FIG. 1 shows a schematic diagram of exemplary general flow patterns in co-rotating and counter-rotating twin screw extruders (two flighted screws in both cases). [Figure 19] FIG. 1 shows a schematic diagram of exemplary shear stress profiles in counter-rotating and co-rotating twin screw extruders. [Figure 20] FIG. 1 shows a schematic diagram of an exemplary possible configuration of a co-rotating twin screw extruder for a multi-process food extrusion system. [Figure 21] FIG. 1 shows a schematic diagram of exemplary extrudate swelling at the die exit. [Figure 22] FIG. 1 shows a diagram that schematically illustrates an exemplary product expansion after exiting the die due to steam pressure differential. [Figure 23] FIG. 1 shows a schematic diagram of exemplary intermeshing orifice plugs (disks) for severe screw restriction and dispersive mixing. [Figure 24] FIG. 1 shows a schematic diagram of an exemplary twin-screw extruder showing simulated pressure and temperature profiles within the extruder. [Figure 25] FIG. 1 shows a schematic diagram of an exemplary screw design with two vent ports for degassing of volatiles. [Figure 26] 1A-1C show schematic diagrams of exemplary vent port adapter designs. [Figure 27]1 shows a schematic diagram of an exemplary extrusion process realized by a typical continuous thermomechanical processing technology that combines several unit operations such as conveying, mixing, shearing, plasticizing, melting, cooking, and polymerization. These multiple unit operations result in a complex correlation between several variable parameters and the respective process responses, thus determining the quality of the product. [Figure 28] FIG. 1 shows a schematic diagram of an exemplary AI- or ML-based automated recipe management and optimization in an inventive extrusion process and in an inventive extrusion system. [Figure 29] 1 shows a diagram that schematically illustrates an exemplary smart recipe selection, which is a process of selecting a suitable recipe from a set of predefined recipes based on a given descriptor of a product. A descriptor is defined as a set of different characteristics of a product. Examples of characteristics are color, fiber content, etc. Such a set can be named (in the context of high moisture extrusion) as "chicken" or "fish." [Figure 30] 1 shows a schematic diagram of an exemplary RDB, which is a database maintained by the system of the present invention that an operator can use to search / download new recipes. The database serves as storage for automatic recipe selection and validates and updates recipe data by inserting recorded process execution insights, which are based on raw material properties such as color, fiber content, etc. and mapped to achievable end product properties. [Figure 31]A diagram showing a schematic of an exemplary IRPC system that generates data on raw materials used in extrusion processes. Its goal is to function as: (a) input data for automatic recipe selection by mapping raw material types to the customer / operator's desired product based on possible achievable product characteristics. It acts as an advisor: the customer / operator can select a specific recipe and the IIC will return whether the desired end product is feasible with the given raw materials, or itself will give suggestions on which raw materials should be used in the end product to achieve the best results, and (b) input for smart process optimization systems via forecasted / projected raw material data. [Figure 32] 1 shows a schematic diagram of an exemplary raw material database for the system of the present invention, including entries for raw product characteristics such as moisture, protein content, etc.; process target information such as CO2 relevance, which may conflict with measurements of final product classification using product characteristics as classifications. [Figure 33] 1 shows a diagram that schematically illustrates an exemplary smart process optimization, which is the act of adjusting the parameters of an extrusion process during production based on given targets: a process target (PT) and a product target (given by the recipe). The optimization process must meet or exceed the targets given by the operator. The process optimizer's optimization is limited by the deviation of the product target, where the target is either the machine recipe or the product characteristics. The process target (PT) is provided in the form of a pre-designed chain of command. [Figure 34] FIG. 1 shows a schematic diagram of an exemplary intelligent end-product classifier that can use sensor information acquired from the end-product from an extruder, for example, to map the product to end-product characteristics. As information sources, the end-product classifier can use, for example, online end-product measurements as well as data from the process itself (temperature, pressure, SME, etc.). Offline end-product measurements can be used to validate the measurements and train the system. [Figure 35] FIG. 1 shows a diagram that schematically illustrates exemplary process targets. A process target is a set of process commands that affect how a process is performed. Possible process targets are: (i) minimizing the energy consumption of the process, (ii) maximizing the throughput of the process, and (iii) minimizing the CO2 emissions. [Figure 36] A diagram showing a schematic of an exemplary automated rework process is shown. Smart Rework Intake is a process that reuses unused product coming out of the extrusion process by sending the product back into the process, which can vary the amount of product reused. The service collects input data from the following sources: the extrusion process itself, and the final product classification, which generates information about the product's condition. [Figure 37] FIG. 1 shows a schematic diagram of an exemplary complete overview with the necessary structure for automated recipe management and optimization of the present invention. [Figure 38] 1 shows a diagram that schematically illustrates an exemplary CO2e monitoring dashboard according to a variant of an embodiment of the system of the present invention. [Figure 39] FIG. 1 shows a diagram that schematically illustrates an exemplary preconditioning process for an exemplary extrusion process. [Figure 40] 3 shows a schematic diagram of an exemplary extrusion process: Preconditioning (see FIG. 39) and extrusion (see FIG. 31) form the basis of the overall extrusion process. DETAILED DESCRIPTION OF THE INVENTION

[0108] 1 to 8 schematically illustrate an architecture for a possible implementation of one embodiment of an industrial extruder system 1 and extrusion process of the present invention. The industrial extruder system 1 comprises a feeder 101, an extruder 102, a forming orifice 103 (die), a collection means 104, and an extruder control unit 500. The feeder 101 supplies a plastically deformable and / or viscous input material 301 to the extruder 102, which continuously presses the input material (301) into and out of the forming orifice (103), forming the output material as extrudate 300, according to the Archimedes screw operating principle. The input material 301 is processed by the extruder 1 by hot extrusion, cold extrusion, warm extrusion, friction extrusion, or microextrusion. The input material 301 extruded by the extruder system 1 can include at least food, metal, polymer, ceramic, concrete, or modeling clay. A collection means 104 collects the extrudate 300 for further processing, such as post-extrusion processing including cutting and / or drying and / or seasoning and / or frying and / or coating, as shown in FIG.

[0109] The extrusion process is controllably steered during operation by an extruder control unit 500, which includes a programmable logic 107 for setting and adapting operational setting parameter values ​​208 of the operating units 105 of the feeder 101, the extruder 102, the forming aperture (die) 103, and the collecting means 104. The programmable logic 107 includes programmable logic controllers of the operating units, for example, a programmable logic controller of a frequency converter for controlling the motor 108. The operational setting parameters 201 include the screw speed of the extruder for pressing the input material 301, and / or the addition rate by the feeder of at least one ingredient for constituting the input material 301, and / or the conditioning settings of the conditioner unit 122 of the extruder 102 and / or the conditioner of the forming aperture 103 for cooling or heating the input material 301 in the extruder 102, and / or the positioning size of the forming aperture 103 area.

[0110] The extruder control section 500 further includes a digital controller 502 for signaling and steering the programmable logic 501. To steer the programmable logic 501, the digital controller 502 receives input parameter values ​​200, including at least process parameter values ​​and / or operational setting parameter values ​​(208) and / or material property parameter values ​​202 and / or environmental measurement parameter values ​​209. The material property parameters 202 include texture, density, color, anisotropy, chemical composition, thickness, degree of polymerization, moisture content, protein content, starch content, fiber content, particle size, surface structure, and / or tolerances. Furthermore, the input parameter values ​​200 include sensory parameter values ​​207 measured by sensors 111 of the feeder 101 and / or extruder 102 and / or molding opening 103 and / or collecting means 104, as shown in FIG. 3 .

[0111] The extruder control 500 includes a repository storage unit 503 having an adaptive digital database 504 holding a plurality of selectable structured data records 507 for storing digital recipes, each of the selectable data records 507 including at least the material property parameters 202 of the input material 301 and the target material property parameters 204 of the extrudate 300 and / or initial operational setting parameters providing an initial setting for the operational setting parameters 201 for operation 105 of the extruder system 1, and the input parameter values ​​200 further including the parameter values ​​of the selected data record 507, as shown in Figure 6. The extruder system 1 includes sending digital signals by the digital controller 502 to operate the programmable logic 501 and associated operational units 507 to controllably and steerably extrude the extrudate 300 having material property parameter values ​​202 within predefined tolerance ranges of the predefined target parameter values ​​206, as shown in Figure 3. 2. The digital database 504 is realized as a digital library, and the repository storage unit 503 has a network interface 506 that provides access to the structured data records 507 via a data transmission network 505 for the selection and / or adaptation and / or creation of the structured data records 507. Data transmission is the transfer and reception of data in the form of a digital bit stream or a digitized analog signal transmitted over a point-to-point or point-to-multipoint communication channel. Examples of such channels are copper wire, optical fiber, wireless communication using the radio spectrum, storage media, and computer buses.

[0112] The digital controller 502 further includes a machine learning unit 510 that monitors and classifies input parameter value 200 patterns and adapts operational setting parameter values ​​208 of the operational units 105 of the feeder 101 and / or the extruder 102 and / or the molding aperture 103 and / or the collecting means 104 to adjust the measured material property parameter value 202 of the extrudate 300 to within a predefined tolerance range. The machine learning unit 510 can be implemented, for example, by including a deep learning (DL) structure including one or more neural network (NN) structures, as shown in FIG. 7 . The one or more neural network (NN) structures can include, for example, an input layer 601, at least one hidden layer 602, and an output layer 603; and / or one or more statistical modeling structures that provide output parameter values ​​213 based on the input parameter values ​​200 indicative of the adaptation of parameter values ​​necessary to adjust the measured material property parameter value 202 of the extrudate 300 to within a predefined tolerance range. A machine learning-based DL architecture includes at least a cascade of multiple layers of nonlinear processing units 512 for feature extraction 512a and signal transformation 512b, with each successive layer using the output of the previous layer as input to provide supervised learning for at least classification and / or unsupervised learning for at least pattern recognition. When mining data, supervised learning is separated into two types of problems: classification and regression: (i) classification problems use algorithms to accurately assign test data to specific categories. In the real world, supervised learning algorithms can be used to classify spam into folders separate from the inbox. Linear classifiers, support vector machines, decision trees, and random forests are all common types of classification algorithms. Preferably, the DL architecture includes at least a convolutional neural network (CNN), a deep neural network architecture, as shown in FIG. 8.

[0113] Regression is another supervised learning method that uses an algorithm to understand the relationship between dependent and independent variables. Regression models are useful for predicting numerical values ​​based on various data points, such as sales forecasts for a given business. Common regression algorithms include linear regression, logistic regression, and polynomial regression. Unsupervised learning uses machine learning algorithms to analyze and cluster unlabeled data sets. These algorithms are "unsupervised" because they discover hidden patterns in the data without human intervention. Unsupervised learning models are used for three main tasks: clustering, association, and dimensionality reduction. (i) Clustering is a data mining technique that groups unlabeled data based on similarities or differences. For example, the K-means clustering algorithm assigns similar data points to groups, with the K value representing the size and granularity of the grouping. This technique is useful for market segmentation and image compression. (ii) Association is another type of unsupervised learning technique that uses different rules to find relationships between variables in a given dataset. These methods are frequently used in market basket analysis and recommendation engines, along the lines of "customers who bought this item also bought" recommendations. (iii) Dimensionality reduction is a learning technique used when the number of features or dimensions in a given dataset is too large. It reduces the number of data inputs to a manageable size while maintaining data integrity. This technique is often used in the data preprocessing stage, such as when an autoencoder removes noise from visual data to improve image quality. Measured and / or captured input parameter value 200 patterns are classified and selected by the convolutional layer 612 and pooling layer 613 of the CNN structure, where the pooling layer 613 reduces the dimensionality of the feature map of the measured and / or captured input parameter value (200) patterns, thus reducing the processing complexity in the machine learning unit 510 to adapt the operational setting parameter values ​​201 of the operational unit 105.

[0114] Faults during the extrusion process are automatically detected by the machine learning unit based on the measured and monitored input parameter values ​​200, and warning and / or steering signals are generated upon detection of a predicted fault in the extrusion process. Preferably, the digital controller 502 includes a fault processing unit that generates an error signal based on the warning and / or steering signals. The operational setting parameters include, in particular, error signals for steering the program logic 501 for setting an error function of each program logic controller of each operational unit 105. The fault processing unit is configurable for a given warning and / or steering signal. Preferably, the digital controller includes a human-machine interface for signaling the warning signal.

[0115] The extruder network 2 includes two or more extruder systems 1 and a central digital controller 514 having a central repository 512 with at least one adaptive central digital database 513 including structured data sets 515 for storing digital recipes and / or ingredients and / or products, wherein at least one of the digital controllers of the plurality of extruder systems 1 is given read and / or write access to the structured central data records 515 via a data transmission network 505 to select and / or adapt and / or generate the structured central data records 515, the central data records generated by the extruder systems 1 having at least one data classification parameter 211 characterizing the generating extruder system or factors affecting the generating extruder system 1. The data classification parameter 211 includes the country of operation of the generating extruder system and / or the operator of the generating extruder system 1 and / or the ID and / or extruder type of the generating extruder system 1 for classifying the data records stored centrally in the at least one central database with data classification.

[0116] I. Consistent product quality and unmanned operation

[0117] Extrusion is a continuous thermomechanical process technology that combines several unit operations, such as conveying, mixing, shearing, plasticizing, melting, cooking, and polymerization. These numerous unit operations lead to complex correlations between several variable parameters and their respective process responses, which determine product quality (see Figure 18). Currently, time-consuming and costly offline analysis, well-trained operators, and laboratory staff are required to assess product quality and adjust the process, and therefore product quality, accordingly. Due to the nature of offline analysis, process adjustments are only made with a time delay. Therefore, online measurements to assess product quality in real time would enable rapid process adjustments and consistent product quality. Furthermore, offline analysis would be reduced, reducing both analytical and staff costs.

[0118] The extruder and its surroundings setup considers fixed parameters such as screw configuration (geometry), barrel length, cooling die geometry, and cooling die insert. Variable parameters include temperature, screw speed, water addition rate, solids addition rate, oil addition rate, and location of liquid addition such as water, oil, flavors, mineral solutions, acids, and caustics. Additionally, nitrogen (N2) introduction as part of aeration techniques is also a variable parameter. Variable parameters determine process responses such as specific mechanical energy (SME), thermal stress (STE), pressure at the end plate, temperature at the end plate, residence time (distribution), weighted average total strain (WATS, Forte), and flow profile / velocity at the cooling die exit. All of the aforementioned process responses are measurable quantities. SME is a quantity calculated from power consumption / torque measurements recorded from the extruder's main drive; STE is calculated based on inlet and outlet temperatures; volumetric flow rate of the heating cycle; pressure and temperature at the end plate are measured by pressure and temperature probes; flow profile and velocity are accessible via an HD camera and underlying video analysis program; residence time (distribution) is measured using an inert indicator, e.g., a colorant, also recorded by an HD camera and tracked over time; and WATS (weighted average total strain) is a quantity calculated using fixed parameters (screw geometry, barrel length, cooling die geometry, and inserts), screw speed, and residence time. In a variation of the embodiment, the ML- or AI-based core engine 11 includes an AI control loop that alters the process response by adapting variable parameters. Furthermore, complex interactions between process response and variable parameters and the product quality of the extrudate can be established. Because changes in a single variable parameter can have more than one effect on process response and therefore product quality without clear correlation, the artificial intelligence realized by this invention is a powerful tool to uncover these interactions, both for routine production and quality assurance, as well as for research and development purposes.An example of this is a prior art system intended to measure and study the effect of SME on the anisotropy of the final product, but it was found that increasing the screw speed to change the SME also increased the temperature at the end plate, and therefore the prior art system could not clearly correlate the observed effect on product quality.

[0119] The product quality of wet extrudates is best described by their texture, anisotropy, density, chemical composition, color, and degree of polymerization. To set up an artificially controlled loop for real-time adjustment of variable parameters, product quality must be analyzed online. The following table (Table 2 is a non-exhaustive list) shows applicable methodologies to enable consistent product quality throughout production. [Table 2]

[0120] If some of the above product qualities cannot be measured directly, a mapping needs to be achieved between measurable process parameters such as temperature, pressure, SME, etc. and the above methods for the above product quality indicators. In a variant embodiment, a machine learning modeling structure is used, for example, to perform the appropriate mapping.

[0121] The resulting final product quality metrics can be used for a number of different applications within the extrusion process, for example, they can be used as validation data for AI applications previously used in the process chain such as automated process optimization to allow operators to define process targets instead of using traditional recipes, allowing good and bad products to be determined, or they can be used as a raw material database where raw materials can be mapped to achievable products.

[0122] II. Automatic Recipe Management and Optimization

[0123] There are various technical problems with automated recipe management and optimization that are solved by the extruder and extrusion system of the present invention: (i) in prior art systems, it is usually unclear whether the extruder's maximum potential has been reached; (ii) in prior art systems, there is little operator know-how and there tends to be little customer know-how; (iii) in prior art systems, recipe optimization is usually very time-consuming and occurs entirely in the engineer's mind, i.e., empirically; (iv) in prior art systems, the optimized product characteristics are usually not clear and must be defined through an empirical approximation process; (v) in prior art systems, recipe management and optimization still needs to be integrated into the extrusion process, i.e., form part of the standard extrusion process. Figure 28 shows a diagram that schematically illustrates an exemplary complete overview with an applicable structure for achieving automated recipe management and optimization of the present invention.

[0124] (a) Smart recipe selection

[0125] Smart recipe selection is the process of selecting a suitable recipe from a set of predefined recipes based on a given descriptor of the product (see Figure 20). A descriptor is defined as a set of different characteristics of the product. Examples of characteristics are color, fiber content, etc. Such a set can be given a name such as "chicken" or "fish" (in the context of high moisture extrusion). Based on the set, the included characteristics can be modified by the operator after selection. For example: the operator changes the color of the final product.

[0126] The recipe selection process is built on a variety of data sources: Recipe database RDB Intelligent raw material classification IIC Intelligent End Product Classification IFPC (described in a later chapter) Process Target PT (described in a later chapter) Operator selection

[0127] Recipe database (RDB)

[0128] The RDB is a database maintained by the invention system that operators can use to search / download new recipes (see Figure 21). The database acts as a storage for automatic recipe selection, and validates and updates recipe data by inserting recorded process execution insights. This is based on raw material properties such as color, fiber content, etc., which are mapped to achievable end product properties.

[0129] Raw material classification (IIC)

[0130] The IRPC system generates data about the raw materials used in the extrusion process. Its goal is to: Input data for automatic recipe selection by mapping raw material types to the customer / operator's desired product based on achievable product characteristics. It acts as an advisor: the customer / operator can select a particular recipe and the IIC will return whether the desired end product is achievable with the given raw materials, or it will itself give suggestions on which raw materials should be used in the end product to get the best result. · Input for smart process optimization systems with predicted / expected raw product data.

[0131] It can be sourced from the following sources: · Ingredients database IDB (see below). Online raw material analysis with in-line devices such as NIR. Offline raw material analysis, e.g. in a laboratory. · Final product classification IFPC, which returns the observed final product characteristics. · The process target selected by the operator. · Operator manually inserts original product information.

[0132] IRPC uses measured online and offline raw product analysis data to detect product deviations. These deviations are stored in the raw material database for training and validation of the IRPC model, as well as input data from the database. IRPC sends predicted raw product characteristics to automated process optimization, thus influencing the process (see also Figure 22).

[0133] Ingredient Database (IDB)

[0134] The raw material database maintained by the system of the present invention includes entries for raw material mass characteristics such as moisture, protein content, etc.; process target information such as CO2 relevance that may conflict with measurements of final product classification using product characteristics as classifications.

[0135] Smart Process Optimization (SPO)

[0136] Smart process optimization is the act of adjusting the parameters of an extrusion process during production based on given targets: process targets (PT) and product targets (given by the recipe). The optimized process must meet or exceed the targets given by the operator. The process optimizer is constrained in its optimization by deviations from the product targets, which can be either machine recipes or product characteristics. The process targets (PT) are provided in the form of a pre-designed chain of command.

[0137] SPO uses the following data sources: · The Intelligent Final Product Classifier (IFPC) returns the measured final product characteristics exiting the process. The extrusion process generates live data from the process, such as product temperature at a given point in the process chain, end plate and (cooling) die pressure, SME, etc. If recipe deviations are used, these values ​​act as deviation limits. · Smart Recipe Selector (SMR) transmits the end product characteristics to be achieved. · Intelligent Ingredient Classifier (IIS) delivers predicted raw material characteristics information. Online / offline product measurement, transmit live / batch raw material information. The operator selects the process target (PT). Smart rework capture to input data about rework that may be fed into the process.

[0138] The Intelligent Process Optimizer can run in two different modes: if the Smart Recipe Selector and End Product Classifier are available, it can run in Characteristic Product Mode, controlling the process based on achieved product characteristics, or if the Smart Recipe Selector is not available or the End Product Classifier is not available, it can run in Actual Data Only Mode, optimizing based on inline measurements.

[0139] Intelligent End Product Classification (IFPC)

[0140] The intelligent end-product classifier uses sensor information acquired from the end-product exiting the extruder to map the product to end-product characteristics. As information sources, the end-product classifier can use, for example, data from the process itself (temperature, pressure, SME, etc.) and online end-product measurements. Offline end-product measurements can be used to validate the measurements and train the system.

[0141] Output data reflecting the characteristics of the final product are reused by multiple systems: Smart process optimizers use this data to adjust processes. Measurements are used as quality indicators seeking to achieve target product characteristics. · A smart rework intake process that uses analysis of product that is not yet usable to analyze how much can be fed back into the process immediately to reduce waste. · An intelligent ingredient classifier that uses the output of the IFPC to map ingredients to achievable end products.

[0142] Process Target (PT)

[0143] A process target is a set of process commands that affect how a process is executed (see Figure 26). Possible process targets are: Minimizing the energy consumption of the process Maximizing process throughput Minimizing CO2 emissions

[0144] The process target serves as input for the following systems: Intelligent ingredient classifier, which influences its advisory function based on the target. For example, if a target CO2 emission is selected, ingredients will be selected based on their CO2 relevance. Smart recipe selector, where recipes are suggested and selected based on achieving specific targets. Smart process optimizers, chain of command influencing the production process itself.

[0145] Automated Rework

[0146] Smart Rework Intake is the process of reusing unused products coming out of the extrusion process by sending the product back into the process, with the possibility to change the amount of product reused (see Figure 27). The service collects input data from the following sources: the extrusion process itself, and the final product classification which generates information about the product's condition. Whether a product can be sent back into the extrusion process depends on the following parameters: Product characteristic values ​​are set as targets, so that products that do not meet the targets can be reused. Product moisture. If the product is too moist, feedback is not possible without product adhesion. · The size of the product chunks. Pieces that are too large cannot be transported.

[0147] Rework intake impacts Smart Process Optimization. Product cannot be understood as the normal "raw material" of a process. Smart Process Optimization receives suggestions for the amount of rework to intake from the Smart Rework Intake Controller.

[0148] III. Cloud-based raw material database for process optimization based on historical data

[0149] Extrusion allows the processing of a wide variety of raw materials, the raw materials and their properties depending on several environmental factors as well as the process steps applied during production, so the raw materials are prone to variations in composition and functionality.

[0150] In the high-moisture extrusion of meat analog products, different protein sources are typically used, including flour (approximately 50% protein), concentrates (60-80% protein), isolates (>80% protein), or even intermediate products from the manufacturing process, such as slurries. In a variant embodiment, to enhance the intelligent extruder system, an ingredient database is created to continuously optimize the extrusion process. In addition to different protein sources and water, formulations may also include carbohydrates (i.e., starch, fiber, sugars), fats / oils, minerals, salts, acids, caustics, and flavors. A comprehensive ingredient database must be constructed that includes the composition and functionality-related properties of each ingredient.

[0151] First, the raw material is analyzed to determine its chemical composition, which is its solid content (moisture content), protein content, fat content, carbohydrate content, and minerals.

[0152] Furthermore, specific knowledge of the protein fraction of a raw material is crucial for determining its potential for subsequent texturization during processing. More specifically, the inter- and intramolecular interactions during texturization depend on the specific amino acid profile. Most interactions between proteins in high-moisture extrusions are non-covalent interactions such as hydrogen bonds, hydrophobic interactions, van der Waals interactions, and electrostatic interactions. Furthermore, sulfur-rich amino acids, especially cysteine, form covalent disulfide bridges, which are the primary interactions for forming thermostable high-moisture extrudates. Therefore, detailed analysis of the raw material is necessary. In addition to knowledge of the amino acid profile, advanced knowledge such as denaturation temperature and kinetics, aggregation temperature, and polymerization kinetics are part of the raw material database. Consequently, the raw material database facilitates recipe development and provides insight into whether the raw material is involved in texturization or acts as an inert filler in the final product.

[0153] Further analysis may be required to identify specific carbohydrates to differentiate between sugars, fiber, and starch. For starch-rich ingredients, it may be important to include knowledge of the amylose-amylopectin ratio and gelation rate, as well as degradation. For sugar-rich ingredients, it is important to include knowledge of melting temperature and caramelization.

[0154] Typically, the raw material database includes raw material properties such as true density, particle size distribution, flowability, wettability, dispersibility, water absorption, ionic strength and pH of the slurry, viscosity, modulus, phase transition from dispersion (viscoelastic liquid) to plasticized mass (viscoelastic solid), and the time, temperature, and strain dependence of all these parameters. The raw material database is a continuously growing database that facilitates the evaluation of new raw materials and the development of new recipes. Furthermore, optimized variable parameters for the extrusion process can be derived.

[0155] Given the above list of necessary fluid and solid properties and the lack of a detailed understanding of cooking, kneading, and transport, it is unlikely that a physics-based model of the complete extrusion process will be readily available. However, there are currently notable attempts to model subsystems of the extrusion process, particularly the die flow of wet-textured proteins.

[0156] Quantifying many of the above properties is a challenge in itself, as laboratory equipment often cannot mimic real-world processes and online measurements are inaccurate due to insufficient or improperly placed sensors available. Often, only pressure and surface temperature are available when viscosity and density values ​​are needed to feed material models.

[0157] Modeling viscoelastic flow in industrial processes with complex, interlocking geometries is extremely challenging and time-consuming, but in principle feasible. The inclusion of complex fluid and semi-solid phase transitions requires innovative simulation techniques, such as meshless methods. This is not currently an industry standard and is an area of ​​ongoing academic research. In the prior art, three software codes are known to be used for similar tasks or capable of capturing the key physics of such processes.

[0158] However, such simplified modeling approaches for extrusion processes have only achieved limited success. From today's perspective, creating material models and measuring the corresponding material properties requires significant resources and time before they can be used industrially. Therefore, data-based methods offer an alternative and more accessible way to link input parameters with resulting quantities and build "models" that ultimately allow for process prediction and optimization.

[0159] IV. Optimized and fully automated boot time

[0160] In today's extrusion process, individual start-up procedures are required by each operator, resulting in wasted raw materials and time (approximately 30 minutes) because the start-up phase is only evaluated objectively. Therefore, the intelligent extruder system of the present invention allows for more accurate and optimized trajectory phases by taking into account measurement quantities for online evaluation of texture (e.g., die cutting force) and process response (e.g., end plate pressure). For example, a pressure probe at the end plate (transition from the extruder barrel to the cooling die) is a suitable quantity to rely on for automating the start-up process. A pressure at the end plate exceeding 2 bar indicates a filled screw section. Therefore, in this embodiment variant, the production phase is automatically initiated.

[0161] Furthermore, online measurement of the slurry during the start-up phase (chemical composition, e.g., by NIR) allows its use as rework. For example, in a variant of the embodiment, the rework is injected into the barrel. In this case, the intelligent extruder system relies on a raw material database to adjust variable parameters, ultimately maintaining consistent product quality during production.

[0162] V. Safe Production

[0163] In the prior art, it is known that plant-based meat analogue products are more susceptible to spoilage and spoilage than meat due to their high protein and moisture content and neutral pH, and therefore require more care. In particular, an effective food safety management system is crucial to ensure safe production.

[0164] Online / Inline CCP Monitoring

[0165] Critical Control Point (CCP): Temperature inside the end plate

[0166] Temperature sensors are installed on the end plates to record the temperature every second. The real-time temperature curve is displayed on the dashboard, which has the following functions: Calculate Log5 reduction conditions based on recipe Set critical values ​​(temperature, duration, etc.) for key food safety parameters based on Log5 reduction criteria Alarm and warning function when temperature falls below critical threshold (CCP error) View an overview of alarms and a list of all time windows / periods when CCP errors occurred Each alarm and CCP error can be selected for further analysis The ability to trace a product back to an error and see what happened to the product and what actions were taken (e.g., equipment cleaned and disinfected, etc.) Generate reports showing CCPs, critical limits, CCP errors, recipes, time frames, operators, corrective actions, etc.

[0167] Online / inline CCP control If a CCP error occurs and no corrective action is taken by the operator, the control system will automatically adjust the process, based on the recipe, to increase the temperature above the critical threshold. A diversion flap is installed at the outlet of the cooling die, and any product produced during the period when the CCP standard is not met is sent to a waste bin.

[0168] Verifying machine cleanliness

[0169] Optical sensors are installed in the preconditioner (see Figure 30), extruder (see Figure 31), cooling die, conveyor, and cutter. After cleaning, optical sensors are used to assess cleanliness (by detecting residues and microbial contamination, etc.), and an alarm system is set up to warn if cleanliness is not up to standard and corrective action is required. For wet cleaning, visual inspection is usually not enough. Microbiological testing must be performed. An ATP test can be recommended (rapid test: 5 minutes). A checkbox is created on the dashboard, and the operator must check whether a microbiological test has been performed or not. Alternatively, machine cleanliness can be determined by an indirect approach (e.g., microbiological testing of downstream products).

[0170] Safety check before starting the extrusion process

[0171] Before starting the extrusion process, the control system performs a self-check procedure: Check cleaning history (e.g. when was the last cleaning) Microbiology checks (e.g., when was the last test) Optical cleanliness check

[0172] Based on the feedback from the above checks, the control system decides whether the extrusion process can begin or if further cleaning / inspection is required.

[0173] VI. Sustainability

[0174] The CO2 equivalent (CO2e) is a standard unit for measuring carbon footprints and is used as a common metric because it takes into account all greenhouse gases and is readily available. It simplifies the issue of climate change and highlights carbon hotspots for targeted action. Furthermore, it complies with ISO standards.

[0175] Raw material database for consulting services

[0176] All relevant raw materials are recorded in a raw material database, including potential suppliers and relevant product specifications, costs and CO2e emission factors. · Able to suggest suitable raw material suppliers based on customer specific requirements. Generate a summary of your CO2e emissions based on your selected raw materials, suppliers, transport, etc. Generate comparisons of CO2e emissions for different raw materials, suppliers and transport.

[0177] CO2e monitoring of processes

[0178] By connecting the system to the Expert Insight system, information on raw materials, throughput and energy consumption is obtained, which can be used to quantify CO2e emissions throughout the process.

[0179] For example, using a particularly suitable CO2e monitoring dashboard (see Figure 29), the following digital monitoring and expert services can be offered: Real-time CO2e emissions throughout the process A breakdown of CO2e emissions for each processing step (allowing you to identify carbon hotspots) CO2e certified by SGS Track sustainability performance over time

[0180] The Expert Insights system can be realized as a central platform for connected products and services, optimizing plant efficiency and reducing maintenance time, energy consumption, and waste. Turning machines into connected devices allows the free flow of data from sensors, machines, and control units to a single, secure storage location. Connecting the plant to the central platform creates a gateway and leverages the benefits of digitalization. The central platform allows for transparency into process and machine data—for example, with individual dashboards displaying the most important KPIs. Such transparency can provide, initiate, or electronically notify specific actions to improve performance and optimize processes. Thus, connected devices, operational metrics, and analytics help optimize plant efficiency. Choose from a wide range of digital services to increase productivity, improve product quality, and reduce waste and energy consumption. Accessed via tablet or smartphone, in a variant implementation, this even makes it possible to control the plant on the move. [Explanation of symbols]

[0181] 1. Extruder System 2. Distributed extruder network system 101 Feeder 102 Extruder 103 Molded opening 104 Collection Methods 105 Operating Unit 111 Sensor 120 screw 121 barrels 122 Conditioning Unit 123 Motor 200 Input Parameters 201 Operational setting parameters 202 Material property parameters 203 Environmental Measurement Parameters 204 Target material property parameters 206 Predefined Target Parameter Values 207 Sensory Parameters 208 Operational setting parameters 209 Environmental Measurement Parameters 211 Initial operation setting parameters 212 Digital Recipes 213 Output Materials 300 extrudates 301 Input materials 302 Output Materials 303 Raw materials 401 Operation Settings 500 Extruder control unit 501 Programmable Logic 502 Digital Controller 503 Repository Storage Units 504 Digital Database 505 Data Transmission Network 506 Network Interface 507 data records 511 Operational Data Storage 512 Nonlinear Processing Unit 512 Central Repository 513 Central Digital Database 514 Central Digital Controller 515 Operating Unit 515 Central Data Record 601 Input Layer 602 Hidden Layer 603 Output Layer 610 Machine Learning Unit 611 Deep Learning (DL) Structure 612 convolutional layers 613 Pooling Layer

Claims

1. 1. An extruder system comprising a feeder, an extruder, a forming orifice (die), a collecting means, and an extruder control, wherein the feeder supplies a plastically deformable and / or viscous input material to the extruder, the extruder continuously forcing the input material into and out of the forming orifice forming an output material as extrudate, the collecting means collecting the extrudate for further processing, the extrusion process being controllably steered during operation by the extruder control, the extruder control having programmable logic for setting and adapting operational setting parameter values ​​of the operating units of the feeder, the extruder, the forming orifice (die), and the collecting means; the extruder control section further comprises a digital controller for sending signals to and steering the programmable logic, wherein to drive the programmable logic, the digital controller takes in input parameter values ​​including at least process parameter values ​​and / or operational setting parameter values ​​and / or material property parameter values ​​and / or environmental measurement parameter values, the input parameter values ​​including sensory parameter values ​​measured by sensors of the feeder and / or the extruder and / or the forming opening and / or the collecting means; the extruder controller has a repository storage unit comprising an adaptive digital database holding a plurality of selectable structured data records for storing digital recipes, each of the selectable data records including at least material property parameters of the input material and target material property parameters of the extrudate and / or initial operational setting parameters providing initial settings for operational setting parameters for operation of the extruder system, the input parameter values ​​further including parameter values ​​of a selected data record; the extruder system including digital signals for steering the programmable logic and associated operating units by the digital controller to controllably and steerably extrude an extrudate having material property parameter values ​​within a predefined tolerance range of a predefined target parameter value; Extruder system.

2. the adaptive digital database is realized as a digital library, and the repository storage unit has a network interface providing access to the structured data records via a data transmission network for selection and / or adaptation and / or generation of the structured data records; 10. The extruder system of claim 1.

3. the digital controller comprises a machine learning unit that monitors and classifies the input parameter value patterns and adapts the operational setting parameter values ​​of the operational units of the feeder and / or the extruder and / or the forming orifice and / or the collecting means to adjust the measured material property parameter values ​​of the extrudate within the predefined tolerance ranges.

3. The extruder system of claim 1 or 2.

4. the machine learning unit comprises at least one deep learning (DL) structure including one or more neural network (NN) structures and / or one or more statistical modeling structures that provide output parameter values ​​based on the input parameter values ​​indicative of a fit of parameter values ​​required to adjust the measured material property parameter values ​​of the extrudate within the predefined tolerance range; 4. The extruder system of claim 3.

5. The machine learning-based deep learning (DL) architecture comprises at least a cascade of multiple layers of non-linear processing units for feature extraction and signal transformation, each successive layer using the output of the previous layer as input to provide supervised learning for at least classification and / or unsupervised learning for at least pattern recognition.

5. The extruder system of claim 3 or 4.

6. the extrusion process is autonomously adapted by the machine learning unit by automatically adapting the operational setting parameter values ​​of the operational units of the feeder and / or the extruder and / or the forming orifice and / or the collecting means to adjust the measured material property parameter values ​​of the extrudate within the predefined tolerance ranges by time-based monitoring of the input measured parameter values.

6. An extruder system according to any one of claims 3 to 5.

7. Faults during the extrusion process are automatically detected by the machine learning unit based on the measured and monitored input parameter values, and a warning signal and / or a steering signal is generated upon detection of a predicted fault in the extrusion process.

7. An extruder system according to any one of claims 3 to 6.

8. The DL structure includes at least a convolutional neural network (CNN) structure as a deep neural network; 8. An extruder system according to any one of claims 4 to 7.

9. The measured and / or captured input parameter value patterns are classified and selected by a convolutional layer and a pooling layer of the CNN structure, and the pooling layer reduces the dimension of a feature map of the measured and / or captured input parameter value patterns, thus reducing the processing complexity within the machine learning unit to adapt the operational setting parameter values ​​of the operational unit.

9. The extruder system of claim 8.

10. the repository storage unit further comprises an operational data storage configured to store historical operational data, the historical operational data including historical input parameter values, historical material property parameter values, and predefined target parameter values, and the machine learning unit is trained by applying the historical operational data.

10. An extruder system according to any one of claims 1 to 9.

11. said material characteristic parameters comprising texture and / or density and / or color and / or anisotropy and / or chemical composition and / or thickness and / or degree of polymerization and / or moisture content and / or protein content and / or starch content and / or fiber content and / or particle size and / or surface structure and / or tolerance range; 11. An extruder system according to any one of claims 1 to 10.

12. the operational setting parameters include a screw speed of the extruder screw for pressing the input material, and / or an addition rate of at least one ingredient constituting the input material by the feeder, and / or a conditioning setting of an extruder conditioner and / or a conditioner of the moulding aperture for cooling or heating the input material in the extruder, and / or a positioning size of the moulding aperture area, 12. An extruder system according to any one of claims 1 to 11.

13. the target parameters include at least one of the material property parameters and / or process parameters including at least the energy consumption of the extruder system; 13. An extruder system according to any one of claims 1 to 12.

14. wherein material parameter values ​​of the input material and / or raw materials of the input material are automatically determined by the machine learning unit, which adapts the input process of the feeder by adapting the operational setting parameters of the feeder.

10. The extruder system of claim 1.

15. 15. A distributed extruder network system comprising two or more extruder systems according to any one of claims 1 to 14 and a central digital controller having a central repository with at least one adaptive central digital database comprising structured data sets for storing digital recipes and / or ingredients and / or products, wherein at least one of the digital controllers of a plurality of said extruder systems is given read and / or write access to said structured central data records via said data transmission network in order to select and / or adapt and / or generate said structured central data records, said central data records generated by said extruder systems comprising at least one data classification parameter characterizing said extruder systems or factors influencing said extruder systems.

16. the data classification parameters include an operating country of the generating extruder system and / or an operator of the generating extruder system and / or an ID of the generating extruder system and / or an extruder type for classifying the centrally stored data records in the at least one central database with the data classification parameters; 16. The distributed extruder network system of claim 15.

Images