Unforeseen downtimes in serial production are always accompanied by high costs, making it essential for machines to be serviced in time. Because sensor, transmission and data storage technology have taken a big leap forward in recent years, predictive maintenance has already been realized in some industries, demonstrating a tangible benefit of Industry 4.0. The WZL and the WBA have developed a guideline for this.

Fig. 1. The guideline for the development of predictive maintenance solutions is divided into three phases: analysis, design and implementation

Guideline to the Development of Predictive Maintenance Systems for Serial Producers and Toolmaking Companies

Industrie 4.0 für die Qualitätssicherung

Industrie 4.0

Instandhaltung

Condition Monitoring / Prozessüberwachung

Werkzeug- und Formenbau

Dienstleistungen

Werkzeug-, Formen- und Modellbau

Sensoren verschiedener Messverfahren

Fertigungsautomation

Sensorik / Aktorik

Fernwartung / Ferndiagnose

Machine Learning

Datenanalyse

Dienstleistungsqualität

Qualitätsmanagement

Industrie 4.0 für die Kunststoffverarbeitung

Instandhaltungsmanagement (TPM)

Vorbeugende Instandhaltung für die Qualitätssicherung

Werkzeugbau

Machine Learning für die Qualitätssicherung

Symbiosis of Toolmaking and  Serial Production

After a year’s forced break of Fakuma, Wittmann Battenfeld is announcing the new servohydraulic SmartPlus machine as the highlight of this year’s trade show. It is claimed to be characterized by high economic and energy efficiency and repeat accuracy.

The new SmartPlus machine is equipped with the HiQ Flow software and the CMS Light condition monitoring system

Spritzgießen

Kunststoffverarbeitung: Verfahren

Schaumspritzgießen

Mikrospritzgießen

Mehrkomponenten-Spritzgießen

Spritzgießmaschinen, allgemein

Kunststoffverarbeitung: Maschinen und Komponenten

Spritzgießmaschinen

Messe Fakuma

kunststoffe-international.com

Spritzprägen

Elastomer-Spritzgießen

Injection Molding Innovations In-House, Live and in Color

Once again, the K 2019 in Düsseldorf, Germany, lived up to its claim to be the leading trade show for plastics technology. At least in the field of injection molding, there were exciting global innovations to see – if you looked closely.

Fig. 1. Smart Operation – the machine is operated with only four buttons (© KraussMaffei)

Injection Molding Trends at K 2019

Smart Factory

Energieeffizienz

Spritzgießwerkzeuge

Werkzeuge für die Kunststoffverarbeitung

Fließpressen

Handling, Handhabungstechnik für die spanende Fertigung

Automatisierungstechnik

Handling, Handhabungstechnik

Spritzgieß-Sonderverfahren

Spritzgießsimulation

Software

Simulation

Linearroboter für die Kunststoffverarbeitung

Industrieroboter

Montagetechnik

Montage-Spritzgießen

Knickarmroboter

Co-Injektion (Barriere-/Sandwich-Spritzgießen)

Hybrid-Spritzgießen (Metall / Kunststoff)

Ressourcenmanagement

Produktion

Technical Excellence thanks  to Digital Assistance

Car windows can be framed more economically and more securely with a novel thermoplastic elastomer based on styrene block copolymers (TPS), as opposed to rubber seals. Advantageously, the typically low viscosity of TPS leads to reduced glass breakage and allows thinner glass to be used. 

Novel Thermoplastic Elastomer with Good Flow Behavior and High Surface Finish 

Kunststoff Werkstoffe

Kautschuk / Elastomere

Messgeräte für Viskosität

Oberflächenveredelung

Flexible Auto-Glass Framing 

The lighting industry is going through a technological change – from light bulb via energy saving lamp to the light emitting diode (LED). Especially for high-power LED the limit of the optical materials is reached, which affects the durability significantly. Transparent liquid silicone rubber offers a promising alternative because of its good thermal properties and high UV-stability.  

LED-Lenses Made of Liquid Silicone Rubber for Light Applications in the Automobile 

Optik

Kunststoffe

Silicone in Automotive Headlights 

The announcements in the field of injection molding give the impression that the developments of recent years are continuing. These include the increasing digitalization of injection molding production, as well as highly integrated production cells that combine very different competencies. The spotlight shines not only on the machinery, but rather on optimum system solutions for integrating molds and individually configured peripheries in the digital and the real world. 

Fakuma 2018: Injection Molding Networked and Integrated 

Steuerungen (Hardware und Software)

ERP-Systeme

Treibmittel

Folientechnik (IMD, IML, FIM, u.a.)

Temperiertechnik

Reinraumsysteme / -technik

The Communicative Machine 

“At our trade show presentation this year, we are concentrating on the automation and digitalization of our injection-molding machines,” says Alfred Schiffer, marketing partner at Dr. Boy GmbH & Co. KG, explaining the activities planned for Fakuma. Three of the four applications at its own tradeshow booth – four more can be seen at its partners – are equipped with Dr. Boy’s own LR 5 linear handling unit.

On the Boy 35 E, the LR 5 compact telescopic handling system is use for a medical LSR application

kunststoffe.de

Individual Automation and Digitalization

Engel Austria GmbH, Schwertberg, Austria has developed a new injection moulding machine for the processing of all common rubber compounds and other elastomer materials. The new Engel flexseal 300 T horizontal machine with a clamping force of 3,000 kN has been designed specifically for the needs of international O-ring and flat gasket manufacturers and sets new standards for the industry with its features. 

Injection Molding Machine for O-Rings and Flat Gaskets 

Flexseal 300 T 

At the Deutsche Kautschuk-Tagung DKT 2015 in Nuremberg, Germany from June 29th to July 2nd, Arburg GmbH + Co KG, Lossburg, Germany, presented a compact manufacturing cell for removal suction cups with a threaded brass insert. The system is centered around a vertical injection molding machine (type: Allrounder 275 V) fitted with an elastomer processing package. 

Insert Molding with EPDM 

Inserttechnik

Automated Elastomer Processing 

Industry 4.0, lightweight construction, and combined technology – that is how one could summarize the current trends and challenges that have been the subject of countless injection molding developments during the past years. In their advance releases for the K exhibition, exhibitors clearly indicate that they are continuing along these lines. 

Injection Molding between Industry 4.0, Lightweight Construction, and Combined Technologies 

Verbundwerkstoffe / Werkstoffverbunde

Hybridtechnik

Trip to a Known Destination 

Despite the proven superiority of high-tech rubber injection molding machine models, robust lean-tech alternatives still have a large body of support, even though the available methods for increasing rubber plastication performance have already been successfully used for over ten years. This is demonstrated with a practical report at the North-German rubber and silicone processor Präzisa in Ratekau near Lübeck. 

The Rubber Injection Molding Machine Manufacturer LWB Steinl Is Staging a Sporty Comparison between Two Injection Molding Concepts 

Steuern, Regeln

Manufacturing Quality Parts Faster 

The Dow Chemical Company (Dow), Midland, USA, has expanded its elastomer technology platforms and capabilities, including a  range of Silastic silicone and fluorosilicone rubber technologies. For example the Silastic LTC 9400 Liquid Silicone Rubber (LSR) product line is a low-temperature-cure (LTC) LSR solution that enables short cure times at low temperatures and features processing benefits including improved rheology for fast low-pressure injection. According to the manufacturer, that creates new opportunities for part design by enabling overmolding of low-melting-point thermoplastics, actives and electrical components. These LSRs also increase robustness by reducing sensitivity to temperature gradients. 

Low-Temperature-Cure Silicone  

Silikone

Extended Elastomer Product Line from Dow 

The integration of individual parts to a multifunctional assembly is one of the important innovation potentials of plastic injection molding, e. g. the production of glass window modules for car rear doors, as practiced at Henniges Automotive. They consist of the glass pane, a web strip, and fastening clips, which are overmolded by a TPE frame to a ready-to-install assembly. For the manufacturing the automotive supplier relies on vertical C-frame injection molding machines from LWB Steinl. 

Assembly of Higher Integrated Car Window Modules through Injection Molding 

KFZ-Zulieferer

From the Vertical Machine into the  Car Door 

LSR optics offers a wide range of applications due to the optical brilliance, the possibility to produce undercut shapes not to mention the competitive pricing. The adhesive surface which is almost impossible to clean however makes the LSR parts vulnerable against contaminations. This characteristic however limits the possible application fields. To overcome this limitation Wilhelm Weber has partnered with Fraunhofer IFAM in a recent research project.

Half of the board was treated with VUV while the other half has covered. Afterwards norm dust was applied onto the board uniformly and than knocked off (once) in upright position. The dust sticks to the untreated areas while falls of the treated areas

Surface Modification of LSR Components Hinders Adhesion of Dirt

LED / OLED-Systeme und Faseroptik

Lichttechnik

Oberflächentechnik

Silikonkautschuk

No Chance for Dust and Contaminations

In the production of plastic parts for the medical and pharmaceutical industry, the requirements for cleanroom environments are becoming increasingly stringent. In this area, not only the demand for all-electric injection molding machines is increasing. There is also an increasing use of electric drives for the implementation of mold movements, which meet the high requirements for cleanliness and reliability.

Fig. 1. Solutions for electrical mold movements, for example a 72+72 cavity stack mold for the closing cone production

Electrical Drives in Injection Molds

Elektrische Antriebe

Antriebe

Mess- / Regel- und Prozessleittechnik

Prozessautomation

Mess-, Steuer- und Regelgeräte

Qualitätssicherung

Qualitätssicherung Kunststoff

Medizintechnik

Medizinprodukte

Mold Technology

Rubber Injection Molding with  Quality Recognition

Automatic Process Monitoring with iQ Clamp Control and Machine Learning

With the increasing digitalization of injection molding, new process parameters are coming under the spotlight. Engel, together with the Polymer Competence Center Leoben and Montan University Leoben, Austria, is using machine-learning methods to explore the potential of mold breathing for quality assurance in rubber injection molding. This research project was supported by the partner company SKF Sealing Solutions Austria.

Technical rubber products, such as gaskets, dampers or connectors, are critical in many applications for the proper and reliable operation of systems, equipment or vehicles. Nevertheless, economic and environmental factors are forcing manufacturers to integrate ever more functions into their components with ever lower resource consumption [1, 2]. Here, manufacturers are supported in continuing to ensure a constantly high quality by the collection and interpretation of process data [3–5].

The iQ clamp control assistance system of Engel Austria GmbH with mold breathing, provides a useful signal that enables automatic clamping force optimization. It also facilitates monitoring of the production process and manual optimization of quality-relevant parameters, such as changeover point and holding pressure time. It was developed for processing thermoplastics, where it is already widely used [6, 7].

As part of a scientific study, the system was evaluated for elastomers. Chemical crosslinking poses a challenge for this application. The goal of trials by the project partners, Engel, Polymer Competence Center Leoben (PCCL) and Montan University Leoben (MUL), was to detect critical process deviations in the manufacture of rubber parts in real time using the mold breathing signal to detect parts out of tolerance already during the manufacturing process without an additional manual quality control step. To simulate process deviations, the mold temperature was deliberately increased during injection molding without adjusting the crosslinking time accordingly. This causes an impermissible change in the dynamic-mechanical properties of the parts, equivalent to producing rejects as a result of process disturbances.

The process monitoring is performed with the aid of PCA-based methods (principal component analysis). These are machine-learning methods and are capable of utilizing clear correlations between different process signals – e. g. mold breathing (Fig. 1). This allows the error identification rate to be significantly improved compared to standard methods, and temperature-related process disturbances to be visualized.

To allow the PCA, and therefore the correlation between the process and quality data in the production process, to be performed, a training dataset must be generated from the process data for acceptable parts. This training set, which is used as a basis for the subsequent inline error identification, is determined experimentally.

Trainings Set as a Basis for Subsequent Inline Error Identification

The tests were performed in Engel’s technical center on an e-victory 740/220 injection molding machine with a maximum clamping force of 2200 kN. For the mold temperature control, heating platens with electrical heating elements were used, of which the setpoint values could be changed by controlling the injection molding machine. A carbon black-filled nitrile-butadiene rubber (NBR, manufacturer: SKF Sealing Solutions Austria) that is typical for industrial applications was processed into sample parts. In preliminary tests, settings were found for all process parameters, which ensure a stable part production process (Table 1). The heating time was chosen so that, at the point of demolding at a mold temperature of 160 °C, the parts have a degree of crosslinking of 90 %.

The settings were kept constant over 20 cycles, in order to generate training data in cycles 5 to 15 after a short transient oscillation. A constant temperature of 163 °C was established, which was measured with the mold temperature sensor (Fig. 2). From cycle 21, the setpoint value of the heating elements was increased to 180 °C, causing the measured maximum mold temperature also to rise. From cycle 52, the control temperature of the heating elements is reduced to 160 °C again.

With the increase of the setpoint temperature, unexpected temperature deviations resulting from possible malfunctions of the heating or thermal elements were simulated. On this basis, the threshold value for impermissible changes of part quality, as well as the response behavior of the process monitoring methods, were investigated.

The dynamic-mechanical properties of the parts are investigated with an electrodynamic test system, type Instron ElectroPuls E3000 (manufacturer: Illinois Tool Works). The test method used can determine the characteristic material behavior in the pressure oscillating load range. The material’s response to the introduced load is determined as a characteristic value of the tangent of the loss angle tan(δ), which describes the ratio between plastic and elastic material behavior. It thereby provides useful information about the degree of crosslinking.

The tan(δ) values determined in the dynamic tests, as expected, correlate during the test with the measured mold temperature from Figure 2. An increase of tan(δ) means an increase of the plastic components, e. g. because of a lower degree of crosslinking  under the changed crosslinking conditions (Fig. 3a).

In the training phase (cycles 5 to 15), tan(δ) has a mean value of 0.209, with a standard deviation (σ) of 0.002. Starting from this, a tolerance range of ±3 σ is defined. If a part has a value outside this range, it is to be considered an unacceptable part. In the time sequence of the test cycles, this range is left as from cycle 23 and only consistently reached again with cycle 58. The object of statistical process monitoring systems – in this case PCA – is now, while the process is running, to use the process parameters to identify whether the part quality lies within the permissible process window. For assessing the investigated monitoring methods, the individual cycles were marked as acceptable or unacceptable parts corresponding to the marked control limits.

Correlation between Mold Breathing and Cavity Pressure

To confirm that mold breathing is suitable for error recognition, the correlation between the breathing signal and the cavity pressure was first investigated in practical tests (Fig. 4). It is conspicuous that the mold breathing signal only shows significant deflections when the mold is almost full, while the cavity pressure rises as soon as the flow front reaches the sensor (Fig . 4a).

This behavior can be expected since, besides the pressure, mold breathing also takes into account the projected, already filled  surface area. At the switchover point, on the other hand, the peaks of both signals are equally pronounced and can therefore be exactly evaluated. If the peaks of the two signals for all cycles are plotted one on top of the other (Fig. 4b), a linear relationship can be clearly seen.

Process Monitoring with Principal Component Analysis

Processes with multiple variables can be suitably monitored using multivariate statistical methods [8, 9]. For the tests, a PCA-based approach (Info Box) was chosen to reliably identify process fluctuations, and therefore quality fluctuations. The recorded process variables, with the exception of the mold temperature, are then processed by principal component analysis, and the SPE (squared predictive error) statistics for each test cycle are computed (Fig. 3b). Cycles 5 to 15 were used as training data. A common, unilateral intervention limit is calculated from these using the PCA method. If the SPE statistics of a cycle lie beyond this limit, the part is identified as probably unacceptable. Here, the common evaluation of the process variables is found to make the change in mold temperature visible in a statistically significant way, without the need to measure it directly. The change of mold temperature is rapidly identified; only from cycle 55 does the system return to normal operation.

The correct classification rate, i. e. the proportion of cycles that are in each case correctly assigned to acceptable or unacceptable parts, is 85 % in these experiments. The basis for decision-making here lies in the violation of the tan(δ)) intervention limits. The 15 % incorrect classifications are all cycles that ought to be unacceptable parts according to PCA, but are acceptable parts according to tan(δ). The monitoring system would thus be oversensitive. However, the incorrectly classified unacceptable parts are underlain by critical process deviations that are correctly recognized by the system and do not have an impact on the quality feature tan(δ) that is specifically investigated, but can bring other quality features out of tolerance.

All the incorrectly classified data are identified with hollow markers (Fig. 3). The process monitoring system thus classifies cycles 3 and 4 as errors (Fig. 3a), which can be attributed to process conditions at the start of the test that have clearly not yet settled to the steady state. Despite the return to normal process conditions from cycle 51, the dynamic properties of the parts only lie within the control limits again after a delay. It is conspicuous that cycles 60 and 62 are qualified as unacceptable parts, although the measured mold temperature is already in the range of normal conditions again at 160 °C. The system thus also responds to other process disturbances. A comparison with Figure 3b shows that exactly these two cycles have higher tan(δ) values than adjacent cycles.

Something similar can be seen with cycle 23, the tan(δ) value of which is lower than that of the adjacent cycles, and also the SPE statistics lie below the intervention limit. Also in the case of acceptable parts, there seems to be a certain relationship between the SPE statistics and part quality according to tan(δ).

In general, the examples show that despite a strong correlation, none of the evaluated process signals is able to represent the crosslinking process absolutely. All the variables recorded in the experiments describe changes of physical properties of the rubber, especially the flow behavior. However, if the properties of the chemical crosslinking are not in a linear relationship with the physical, the part properties that are primarily determined by the degree of crosslinking cannot be clearly predicted. If all the process signals as well as the multivariate monitoring indicate that the flow properties have changed, even if the predicted tan(δ), which is determined principally by the curing reaction, lies within the intervention limits, then detailed part checking should be performed. Only then can it be assessed whether the unexpected process fluctuations are uncritical or whether they lead to rejects.

Based on the iQ clamp control intelligent assistance system, it was possible, using principal component analysis, to set up a multivariate process monitoring system for rubber injection molding that immediately identifies changes in the process conditions, and classifies the corresponding molded parts as unacceptable. The advantages of this system lie in the low training effort and the possibility of simultaneously processing multiple process variables. Additional information about the significantly better performance compared with standard methods can be requested from the authors. The challenge in elastomer injection molding consists in the fact that most available process signals correlate strongly with the flow properties of the materials, however only weakly with the chemical reactivity, which further sets limits on process signal-based fault identification. The system developed by Engel, PCCL and MUL shows trends in process fluctuations even before they have affected the part quality. This opens up great potential for optimizing series processes in the rubber-processing industry.

[1] Lindsay, J. A.: Practical guide to rubber injection moulding. Smithers Rapra, Shawbury, Shrewsbury, Shropshire 2012

[2] Fasching, M.; Friesenbichler, W.; Pacher, G. A.; Berger, G.: Advanced Mold Concepts to Improve Part Quality and Productivity in Rubber Injection Molding. In: Proceedings of the Regional Conference of the Polymer Processing Society, Graz 2015, pp. 49

[3] Chen, Z.: Data-Driven Fault Detection for Industrial Processes: Canonical Correlation Analysis and Projection Based Methods. Springer Fachmedien, Wiesbaden 2017

[4] Yang, Y.; Chen, X.; Lu, N.: Injection molding process control, monitoring, and optimization. Hanser, München 2016

[5] Zeppenfeld, M.; Hirn, R.; Wardenberg, M.: Wireless Life Cycle Monitoring in the Injection Molded Part. Kunststoffe international 109 (2019) 3, pp. 11–16

[6] Pillwein, G.; Lettner, J.; Giessauf, J.; Steinbichler, G.: Breathing Exercises for Process Optimization. Kunststoffe international 105 (2015) 10, pp. 81–84

[7] Pillwein, G.; Lettner, J.; Lehr, M.: Using Breathing Signals for Process Optimization. Kunststoffe international 110 (2020) 1, pp. 26–31

[8] Hutterer, T.; Berger-Weber, G. R.; Friesenbichler, W.: Fault detection in rubber injection molding with multivariate statistics. Rubber World (2019), pp. 40–44

[9] Hutterer, T.; Berger‐Weber, G. R.; Kerschbaumer, R. C.; Friesenbichler, W.: Rubber injection molding: Applying multivariate statistics to identify quality issues solely from process signals. Polymer Engineering & Science (2020), <a href="https://onlinelibrary.wiley.com/doi/full/10.1002/pen.25604">https://onlinelibrary.wiley.com/doi/full/10.1002/pen.25604</a>

[10] Yang, Y.; Gao, F.: Injection molding product weight: Online prediction and control based on a nonlinear principal component regression model. Polymer Engineering & Science 46 (2006) 4, pp. 540–548

[11] Wold, S.; Geladi, P.; Esbensen, K.; Öhman, J.: Multi-Way Principal Components and PLS-Analysis. Journal of Chemometrics (1987) 1, pp. 41–56

[12] Russell, E. L.; Chiang, L. H.; Braatz, R. D.: Data-driven Methods for Fault Detection and Diagnosis in Chemical Processes. Springer, London 2000. URL: <a href="http://dx.doi.org/10.1007/978">http://dx.doi.org/10.1007/978</a>–1–4471–0409–4

[13] Yang, Y.; Gao, F.: Injection molding product weight: Online prediction and control based on a nonlinear principal component regression model. Polymer Engineering & Science 46 (2006) 4, pp. 540–548

Engel Austria GmbH  

https://www.engelglobal.com

Montanuniversität Leoben  Inst.f. Kunststoffverarbeitung

https://www.unileoben.ac.at

PCCL - Polymer Competence Center Leoben GmbH 

https://www.pccl.at

Montanuniversität Leoben Dept. Kunststofftechnik 

https://www.kunststofftechnik.at

Ausgabenartikel

The collection and interpretation of process data supports producers of safety-relevant rubber products in ensuring a constant high quality

Table 1. In preliminary tests, settings ensuring a stable manufacturing process were found for all process parameters. These parameters were held constant in all experiments

Fig. 1. Mold breathing: The mold is compressed by the build-up of the clamping force. During filling, the resulting part generates an opening force, which is counteracted by the compression. The opening force is proportional to the cavity pressure and to the projected surface area 

Fig. 2. The maximum mold temperature at the cycle end follows immediately after the setpoint values of the heating elements

Fig. 3. The tan(δ) values of the elastomer parts are strongly influenced by the mold temperature (left). SPE statistics allow changes in the process conditions potentially leading to rejects to be recognized (right) [9]

Fig. 4. Mold breathing responds with a delay compared to the cavity pressure, but then simultaneously reaches a pronounced peak (left). Maxima of mold breathing and cavity pressure are strongly positively correlated (right)

Name - Hauptartikel

Principal component analysis is a dimensionality reduction method for increasing the defect recognition rate during process monitoring. At the same time, the complexity can also be reduced. Its particular suitability for industrial processes lies in the fact that it can eliminate linear dependencies between process signals. More detailed insights into the methodology and other application reports can be found in Russel, Chiang et al., Yang, Chen, et. al. and others [4, 10 – 13].

DI Dr. mont. Thomas Hutterer is a research assistant at the Polymer Competence Center Leoben (PCCL), Leoben, Austria; <a href="mailto:thomas.hutterer@pccl.at">thomas.hutterer@pccl.at</a>

Assoz. Prof. DI Dr. mont. Gerald R. Berger-Weber is a research assistant at the chair of injection molding of plastics at Montan University Leoben; <a href="mailto:gerald.berger@unileoben.ac.at">gerald.berger@unileoben.ac.at</a>

DI Dr. mont. Roman C. Kerschbaumer is group manager for processing technology at the PCCL; <a href="mailto:roman.kerschbaumer@pccl.at">roman.kerschbaumer@pccl.at</a>

Univ.-Prof. DI Dr. mont. Walter Friesenbichler is Head of the Chair of Injection Molding of Plastics at Montan University Leoben; <a href="mailto:walter.friesenbichler@unileoben.ac.at">walter.friesenbichler@unileoben.ac.at</a>

DI Alexander Klutz is development engineer in Global Application Engineering at Engel Austria GmbH Schwertberg, Austria.

DI Thomas Leng is Vice President Global Application Engineering at Engel; <a href="mailto:thomas.leng@engel.at">thomas.leng@engel.at</a>