1 Introduction

With the development of Industry 4.0, the digitalization of the manufacturing industry has become a topic of global concern, especially for the injection molding industry, which requires low costs, automation, high production efficiency, competitiveness, and safe human–machine collaboration to achieve smart manufacturing. The core of smart manufacturing is digital (information) technology that automatically collects and analyzes data during the product manufacturing process. Given the continuous accumulation of manufacturing data, the construction of cyber-physical systems with artificial intelligence technology can facilitate sensible decision-making and optimize processes. In the field of injection molding, although powerful mold flow simulation software has assisted in optimizing product design, mold design, and part quality evaluation, simulations for determining the optimal process parameters require high accuracy before the parameters can be applied to molding. Although applying sensing technology to tasks such as identifying cavity pressure curves to determine the physical flow behavior of polymer melts can help optimize process parameters during molding, the cost and feasibility of sensors limit their use, thus necessitating simulations. However, because of a lack of accurate information regarding the machines, molds, processed resins, auxiliary devices, environmental noise, and human factors that affect the quality of molded parts, simulated molding differs considerably from real molding. This may be due to differences in injection speed response time between simulations and real scenarios. In particular, the heat transfer coefficient, viscosity, and compressibility of the processed resins under various pressure and temperature conditions may be the main variables for discrepancies between the molding and simulation results. Other sources of error, such as check ring leakage, can also cause discrepancies. Mold rigidity can reduce pressure attenuation in the holding stage considerably, especially for thin-walled molding, in which the pressure curve controls the residual stress acting on the polymer melt. In addition, in the Williams–Landel–Ferry model, which is widely used in injection molding simulation software because of its simplicity, a considerable difference is evident between the pressure predicted by CAE molding simulations and the pressure observed by real molding operators, resulting in the selection of inappropriate machines for injection molding or erroneous mold design.

The injection molding process parameters that most strongly affect molding quality are melt temperature, cooling temperature, injection speed, the velocity-to-pressure (V/P) switchover point, holding pressure and time, clamping force, and back pressure. These process parameters can be systematically optimized using sensor technology to determine the cavity pressure, screw position, and melt temperature of a polymer melt in the injection molding process. To consistently produce high-quality parts, the quality of injection molding must be monitored in addition to the quality of the process parameters. Cavity pressure is an indicator of molding quality. However, sensing is costly and requires installation space, and simulations must be accurate when the pressure curve is applied to quality control. Therefore, eliminating the difference between the results of simulated and real molding is essential and represents the purpose of this research.

2 Literature review

Injection molding is a pressure-driven process. Polymer melts require pressure to overcome resistance when flowing through nozzles, sprues, runners, and gates and entering cavities. Therefore, the system pressure determines the quality of injection molding. The cavity pressure curve can indicate the quality of injection-molded parts (e.g., short shot and flash), and by examining the slope of the cooling phase, the shrinkage rate of parts can be determined [1]. The cavity pressure curve is usually used to monitor the quality of repeated injections. However, repeated injections with the same machine parameter settings may produce different cavity pressure curves at the end of the flow. This indicates that high-precision injection molding machines do not guarantee consistent injection molding quality.

Various sensing technology methods can be used to determine the flow behavior of polymer melts in a mold. Therefore, the injection molding process no longer relies on engineers’ experience to determine the process parameter settings that ensure high quality and yield. Instead, the process requires sensor data provided by an injection molding machine and several sensors installed in the mold or machine. Sensors can be divided into in-mold sensors, nozzle sensors, and tie-bar sensors depending on their installation position. In-mold sensors include pressure and temperature sensors, which are used to determine the flow behavior of polymer melts in sprue–runner–gate and cavity systems [2,3,4,5,6,7]. Nozzle sensors are used to identify changes in the viscosity of polymer melt as it flows from a barrel and passes through a nozzle, which helps to monitor the quality of plasticization before injection into the mold [8,9,10,11,12]. Tie-bar sensors include strain gauges and ultrasonic sensors. They are installed in the machine and are nondestructive to the mold, which is why they are also called “nondestructive sensors.” Although they are not as sensitive a sensor as cavity pressure, they are useful in quality monitoring and control [13,14,15,16,17,18,19,20].

Beaumont [21] studied a Therma-flo method that accounts for the effect of part thickness and melt and mold temperature on polymer melt rheology and concluded that for certain thick parts and fast injection speeds, melt temperature can be balanced through adjustments to shear and thermodynamic heat. Hopmann and Zhuang [22] determined the injection pressures required for various filling time settings and identified an optimal injection speed that produces the least resistance during the filling process. The optimal injection speed also resulted in a minimal change in viscosity, helping to achieve consistent quality in mass production [23]. Karbasi [24] revealed that temperature affects pressure distribution, especially in the plasticizing phase and V/P switchover point. Nian et al. [25] used in-mold machine signals to set the ideal injection speed, injection speed segment, V/P switchover point, and holding conditions on the basis of screw position and cavity pressure. Similarly, Chang et al. [26] applied infrared temperature and cavity pressure sensors to detect melt temperature and pressure and adjusted the holding conditions to minimize part shrinkage on the basis of pressure–volume–temperature (pvT) theory.

With Industry 4.0, cyber-physical systems have become the primary method of intelligent injecting molding, in which mold flow simulation is crucial; eliminating the difference between the results of simulations and real scenarios is the main purpose. Simulation accuracy is limited when simplified mathematical models are employed and is also limited by material property settings, process parameter conditions, and machine performance and aging. Guerrier et al. [27] and Regi et al. [28] regarded the modeling of the machine nozzle and barrel as a hot runner system. The new model increased the consistency of filling times and filling patterns between simulations and real molding. The amount of polymer melt compression in the filling phase must be considered in a simulation. Huang et al. [29, 30] indicated that a machine’s response to the speed command affects the simulation of the filling phase. By calibrating the speed response, the simulation accuracy for speed and pressure can be increased. However, additional factors must be evaluated to further increase simulation accuracy [31].

This study developed a process parameter adjustment technology to eliminate the difference between the results of simulation and real molding. An in-mold machine molding trial method is proposed herein for determining the optimal process parameter settings on the basis of pressure and screw position curves. The process parameters for the simulation, namely, filling stroke, injection speed, and filling-to-packing V/P switchover point, are adjusted on the basis of the ideal pressure and screw position observed during real molding. The process is as follows:

  1. 1.

    Calibrating the full volumetric filling point: by adjusting the filling stroke of the screw position, consistency between the simulation and real molding can be achieved. Consistency in flow behavior helps to adjust the difference in pressure. This process involves complex compression of the polymer melt, which is often inaccurately predicted by simulations.

  2. 2.

    Calibrating the system pressure curve: the injection speed in the simulation is adjusted on the basis of the results of step 1. The system pressure reflects the history of the polymer melt in terms of the system pressure required to overcome resistance when flowing through nozzles, sprues, runners, gates, and cavities. With the history of the screw position, system pressure can be used to determine the change in viscosity and energy working on the polymer melt during the molding process.

  3. 3.

    Calibrating the cavity pressure curve: the V/P switchover point in the simulation is adjusted on the basis of the results of step 2 to ensure the simulated cavity pressure curve is consistent with that observed during real molding. Cavity pressure reflects the history of a polymer melt’s behavior in the cavity during filling, holding, and cooling. The cavity pressure curve indicates injection molding quality and has been used to determine the optimal process parameter settings and predict the quality of injection-molded parts.

  4. 4.

    Quality comparison: the width reduction rate in the simulation after process parameter adjustment is compared with that observed during real molding.

To determine the feasibility of this method, this study examined injection-molded flat plates and compared the simulated pressure and screw position curves before and after calibration with those observed during real molding.

3 Methodology

3.1 In-mold and machine sensing-based molding trial method

The molding trial method based on in-mold machine sensing involves using the information regarding the pressure of the polymer melt in the cavity and the system pressure and screw position of the injection molding machine to identify the ideal process parameters. Screw position and pressure history during injection molding can be used to distinguish the filling, packing, V/P switchover, and cooling phases. The optimal process parameters can be quickly and accurately obtained by using the cavity pressure curve. In injection molding machine production, using the curve as the standard cavity pressure curve can reduce the time required for debugging and setting the process parameters.

The process comprises three steps:

  1. 1.

    Optimizing injection speed during the filling stage: during the filling phase, the injection molding machine performs speed control and employs an upper-limit pressure strategy for the injection screw, which pushes the polymer melt into the cavity. The optimal injection speed is determined on the basis of the minimum pressure drop between the near-gate area and the end of the filling area under various injection speed settings;

  2. 2.

    Optimizing the V/P switchover point: by observing the screw position and pressure history curve, early, late, and ideal timing for switching between the speed control of the injection screw and the pressure control during the mold filling and packing phases can be determined. With early timing, pressure decreases considerably then increases at the switching point. With late timing, pressure peaks. Ideal timing produces a stable pressure curve [25]; and

  3. 3.

    Optimizing the holding pressure and time during the holding phase: by observing the screw position and hold time, the minimum packing time sufficient to compensate for plastic shrinkage can be determined. Several holding conditions are used to determine the process parameters to produce high-quality parts. With the four pressure sensors installed along the filling path (\({A}_{1}\) to \({A}_{4}\)) as an example, the steps for setting multistage holding pressure and time are as follows [32]:

    1. (a)

      Determine the effective holding time—the solidification time of the gate. After the gate is solidified, the melt no longer enters the cavity, and the pressure curve near the gate is no longer affected by the injection pressure outside the gate, allowing the effective holding time to be determined. The holding pressure is initially set to zero then changed to a higher holding pressure (\({P}_{pulse}\)); the pressure curve close to the gate can then be observed. The effective hold time (\({t}_{\Sigma ,hold}\)) represents the minimum holding time during which the pressure curve close to the gate is not affected by \({P}_{pulse}\);

    2. (b)

      Set the holding pressure time for each stage. Similar to (a), the holding pressure is initially set to zero and then increased (\({P}_{pulse}\)). The holding pressure in the first stage of this case study was set to fully compensate for the volumetric shrinkage of the molded part in the area far from the gate. The first-stage holding time (\({t}_{{S}_{4}}\)) in Fig. 1 represents the minimum holding time during which the pressure curve close to \({A}_{4}\) is not affected by \({P}_{pulse}\). The second-stage (\({t}_{{S}_{2}}\)), third-stage (\({t}_{{S}_{3}}\)), and fourth-stage holding times (\({t}_{{S}_{4}}\)) are determined using \({A}_{3}\), \({A}_{2}\), and \({A}_{1}\), where \({t}_{S2}={t}_{{A}_{3}}-{t}_{{A}_{4}}, {t}_{S3}={t}_{{A}_{2}}-{t}_{{A}_{3}}, \mathrm{and }{t}_{S4}={t}_{\Sigma ,hold}-{t}_{{A}_{2}}\); and

    3. (c)

      Set the holding pressure for each stage. The first-stage holding pressure is the same as the single-stage holding pressure to ensure it compensates for the quality of the plastic parts at the far gate (\({A}_{4}\)). The second-stage holding pressure is then reduced. At each induction position, the pressure curve close to the gate must be greater than the pressure curve at the far gate to prevent the melt from flowing back into the cavity; hence, the cavity pressure curve should follow \({P}_{{A}_{1}}>{P}_{{A}_{2}}>{P}_{{A}_{3}}>{P}_{{A}_{4}}\).

Fig. 1
figure 1

Sensed pressure curves at positions \({A}_{1}\) to \({A}_{4}\) with appropriate V/P switchover point and multistage holding process at mold filling and holding stages [32]

Full size image

3.2 Process parameter adjustment in simulation

Because the cavity pressure and screw position curves are determined by factors such as injection speed, V/P switchover, and holding pressure, this study evaluated adjusting these factors to calibrate the simulated pressure curves. First, this study referred to the real system pressure history and adjusted the metering stroke and injection speed in the simulation to ensure consistency in peak pressure, the foremost position of the injection screw, and the V/P switchover timing between the simulation and real molding. This study also referred to the cavity pressure history and adjusted the V/P switchover point to increase consistency.

4 Experimental setup

Figure 2 presents a flat plate with a length of 200 mm, a width of 60 mm, and thicknesses of 1.5, 1, and 2 mm along the flow channel. The plate has a high ratio of flow length to thickness (more than 150). The middle of the plate has a rectangular hole. The diameter of the sprue of the injection mold is 3.5–6 mm, and the thickness of the fan gate is 2–1.5 mm. The cooling system consists of four linear channels with a diameter of 8 mm, and two channels each are used for the male and female mold plates. The processed polymer material was acrylonitrile butadiene styrene, which is an amorphous material (PA756, Chi-Mei Corporation, Taiwan). Figure 2 also displays the measurement position of the part width, where W1, W2, and W3 represent the injection molding quality of the 1.5-mm-thick area and W4 and W5 represent the quality of the 1- and 2-mm-thick areas, respectively. The width was measured with a three-coordinate measuring machine (CRYSTA-Apex S 7106, Mitutoyo Corporation, Japan). This study used a high-precision all-electric injection molding machine (S2000i100B, Fanuc Corporation, Japan) with a maximum clamping force of 1000 kN, a screw diameter of 28 mm, a maximum injection pressure of 240 MPa, and a maximum injection speed of 500 mm/s. The screw stroke was 95 mm. An oil-heating mold temperature control device (YBMI-200–20, Taiwan Yann Bang Electrical Machinery Co., Ltd., Taiwan) was used for the experiment.

Fig. 2
figure 2

Injection-molded part: geometric dimensions and measuring position for width

Full size image

Figure 3 displays the positions of the eight pressure sensors installed in the injection mold. P1, P2, and P3 measure the pressure history at the bottom of the sprue and the front and back of the fan gate, respectively. P4 and P5 measure the pressure history when the 1.5-mm-thick area is filled. P6 and P8 measure the pressure history when the 1-mm-thick area is filled. P7 measures the pressure history when the 2-mm-thick area is filled. This study used two types of pressure sensors (SSB04KN10 × 08H and SSB16KN12 × 10H, Futaba Corporation, Japan). The first was used to measure the cavity pressure, and the second was used to measure the pressure in the sprue–runner–gate system. Figure 3 also displays the position of the three thermocouple sensors installed on the female mold plate (KEX-H-7/0.3 × 2, Championtech Technology Co., Ltd., Taiwan). T1, T2, and T3 were used to measure the mold temperature history. The pressure, screw position, and temperature signals were collected using a data acquisition card (USB-6343 DAQ, National Instruments Co., USA).

Fig. 3
figure 3

Sensing position: pressure (P1 − P7) and temperature (T1 − T2)

Full size image

For the mold flow simulation, this study used Moldex3D 2020 Studio (R1OR version, CoreTech System Co., Ltd., Taiwan). This version provides a three-dimensional simulation of polymer melt compression in the barrel, allowing injection pressure to be accurately predicted [33]. This study used a model of a full mold base with a hybrid mesh for the mold flow simulation. The total number of mesh elements was approximately 10 million. Figure 4 displays the consistency between the simulation and real molding under various mold temperatures at T1, T2, and T3. When the mold temperature was 60 °C, the difference between the simulation and the real molding was less than 1 °C.

Fig. 4
figure 4

Mold temperature curves (dashed line: simulation; solid line: real molding)

Full size image

5 Results and discussion

5.1 Real molding

The real molding process followed the molding trial process based on in-mold machine sensing [25]. The first step was to determine the optimal injection speed on the basis of the minimal pressure drop along the filling path and to determine the injection speed segmentation that would ensure consistent melt-front speeds in various cross-sectional areas. Table 1 presents the process parameter settings in the real molding experiment. Figure 5a presents the system pressure drop under various injection speed settings. When the injection speed was 60 − 70 mm/s, the minimum pressure drop was observed along the flow path of P3 and P5. Therefore, the optimal speed was determined to be 70 mm/s, which allows for the lowest flow resistance. The simulation results are consistent with the real molding results (Fig. 5b).

Table 1 Process parameter settings (real molding)
Full size table
Fig. 5
figure 5

Pressure drop under different injection speed settings: a real molding; b simulation

Full size image

To determine the appropriate V/P switchover point (usually 95 − 98% cavity volume filling), the full volumetric filling stroke of the injection screw during the mold filling process must first be determined. Figure 6 presents the cavity pressure and screw position curves during full volumetric filling. By checking the starting signal at P3 (near the gate) and the volumetrically filled point signal at P8 (at the end of the filling path), the full volumetric filling stroke of the screw can be identified. The full volumetric filling stroke of the screw in the real molding and simulation was 26.94 and 27.20 mm, respectively. The difference was only 0.26 mm (approximately 0.96%), which means that the simulation and real molding were consistent. However, the position of the screws on the front differed (approximately 8 and 16 mm for the real molding and simulation, respectively). Unlike in real molding, rebound movement was not observed immediately after the V/P switchover in the simulation. The pressure on the polymer melt after compression may have differed; this was evaluated in the subsequent experiment.

Fig. 6
figure 6

Cavity pressure and screw position curves during full volumetric filling: a real molding; b simulation

Full size image

Having the holding time correspond to the holding pressure is essential for the effect of postfilling (i.e., holding) and determines the geometric accuracy and the inner stress of a part. However, extending the holding time after the gate is frozen is time-consuming and ineffective. The effective holding time can be determined by identifying the point at which the cavity pressure curve becomes unaffected by an increase in holding time. The minimum time is called the “effective holding time.” Fig. 7 presents the cavity pressure curves of the near-gate P3 under various holding time settings. The holding time for the real molding and simulation was 5 and 3 s, respectively: a considerable difference (approximately 40%). The real temperature of the polymer melt after it left the nozzle may have differed from that in the simulation, resulting in a difference in heat transfer behavior at the gate. Other differences may have been due to the simplified viscoelastic model used to calculate the shear heat dissipation, which should be investigated in future research.

Fig. 7
figure 7

P3 cavity pressure curves under various holding time settings: a real molding; b simulation

Full size image

Regarding the optimization of the holding pressure and time during the holding phase, Fig. 8 presents the cavity pressure curves after the multistage holding conditions were set. On the basis of pvT theory, the residual pressures of P3 − P8, representing the position of the polymer melt near the gate, in the middle of the filling path and away from the gate cavity decreased to less than 10 MPa after 7 s, thereby reducing the nonuniform volume at each position before mold opening. This process helped to effectively reduce the change in shrinkage, thereby reducing the warpage of the injection-molded parts. Table 2 presents the widths of the injection-molded parts under single-stage and multistage holding conditions. The volumetric shrinkage rate under the single-stage holding setting was large, which led to considerable warpage of the parts. By contrast, when the multistage holding setting was used, the shrinkage rate decreased to 0.45% (originally 1.18%), thereby controlling the warpage of the part.

Fig. 8
figure 8

Cavity pressure curves under multistage holding pressure setting (real molding)

Full size image
Table 2 Quality comparison between single-stage and multistage holding settings
Full size table

5.2 Adjusting the simulation process parameters on the basis of the real system pressure

5.2.1 Before calibration

Figure 9 presents a comparison between the simulated and real system pressure and screw position curves under the same process parameters. The difference in peak system pressure was 47% (real molding: 163 MPa; simulation: 240 MPa). After the V/P switchover point, the system pressure in the simulation reached the maximum (240 MPa), indicating a considerable difference between the simulation and real molding (163 MPa). When the polymer melt entered the cavity from S1 (see Fig. 3), the difference in the melt front position was substantial. Although the simulation and real molding yielded the same screw position curve in the initial filling stage, the difference in the foremost position of the screw was 83% (real molding: 6 mm; simulation: 11 mm). In addition, the simulation failed to predict the rebound behavior of the injection screw observed in the real molding (1.3 s [Fig. 9]).

Fig. 9
figure 9

System pressure and screw position curves (dashed line: simulation; solid line: real molding)

Full size image

5.2.2 Calibrating the full volumetric-filling point with metering adjustment

Because of the inconsistency between the simulation and real molding in terms of the foremost position of the injection screw (difference of 4.67 mm), the metering position of the screw in the simulation was adjusted to 48.3 mm (originally 52.97 mm; Table 1). The real molding and simulation were expected to yield the same foremost position for the injection screw. Figure 10 presents a comparison of the simulated and real system pressure and screw position curves after metering adjustment was performed in the simulation. When the polymer melt entered the cavity from the S1 position, the difference in the position of the melt front decreased. Although a considerable difference in the metering stroke setting (real molding: 52.97 mm; simulation: 44.30 mm) was evident, the foremost position of the screw was consistent, and the rebound behavior of the screw was observed in the simulation. The simulated peak pressure (175 MPa) was similar to that of the real molding (163 MPa).

Fig. 10
figure 10

System pressure and screw position curves (dashed line: simulation after metering adjustment; solid line: real molding)

Full size image

5.2.3 Calibrating the system pressure curve with injection speed adjustment

To ensure the consistency in the system’s peak pressure time between the simulation and real molding, this study adjusted the first-stage injection speed to 70 mm/s (originally 84 mm/s; Table 3). Figure 11 displays the consistency between the simulation and real molding in terms of the system pressure curves. The simulated peak pressure (171 MPa) was similar to that of the real molding (163 MPa). However, the filling time of the full cavity volume in the simulation was lower than that of the real molding, indicating differences in the compression of the polymer melt.

Table 3 Process parameter settings in the simulation (adjusted metering and injection speed)
Full size table
Fig. 11
figure 11

System pressure and screw position curves (dash line: simulation after metering and injection speed adjustment; solid line: solid line)

Full size image

This study examined the various cavity pressure curves in the simulation after the metering and injection speeds were adjusted. Figure 12 presents a comparison of the cavity pressure curves between the simulation and real molding. The P2 and P3 cavity pressure curves were similar between the simulation and real molding (Fig. 12a, b). However, the peak values of P2 and P3 in the simulation were considerably lower than those of the real molding, indicating that the flow resistance at the runner and gate may have been underestimated in the simulation. In the real molding, the peak value of P7 was higher than that in the simulation, indicating an error in the pressure simulation along the filling path. P7 on the slope during the holding and cooling stages was also inaccurate in the simulation. For the P8 cavity pressure curve, the simulation differed substantially from the real molding.

Fig. 12
figure 12

Cavity pressure curves (dash line: simulation after metering and injection speed adjustment; solid line: real molding)

Full size image

After the metering stroke and injection speed adjustment, the difference in the metering stroke was 16.3% (real molding: 52.97 mm; simulation: 44.3 mm). The difference in the foremost position of the screw was 0.8% (real molding: 6.18 mm; simulation: 6.13 mm). The difference in the system’s peak pressure was 4.9% (real molding: 163 MPa; simulation: 171 MPa). The difference in peak P3 pressure was 1.4% (real molding: 70 MPa; simulation: 71 MPa). The difference in peak P8 pressure was 93.8% (real molding: 32 MPa; simulation: 2 MPa).

5.2.4 Calibrating the cavity pressure curve with V/P switchover point adjustment

To ensure the simulated P8 cavity pressure curves were consistent, this study adjusted the V/P switchover point to 5.4 mm (originally 6.2 mm). The real peak system pressure was lower than that in the simulation, and the foremost positions of the screw were inconsistent. The P1 and P2 cavity pressure curves were similar between the simulation and real molding. However, the peak values of P1 (Fig. 13a) and P2 (Fig. 13b) in the simulation differed from those in the real molding, suggesting that the flow resistance at the runner and gate may have been underestimated in the simulation. In the real molding, the peak value for P3 (Fig. 13c) was lower than that in the simulation, indicating an error in the pressure simulation along the filling path. The simulation of P3 on the slope during the holding and cooling stages was accurate. In terms of the P8 cavity pressure curve (Fig. 13d), the simulation was highly consistent with the real molding.

Fig. 13
figure 13

Cavity pressure curves (dash line: simulation after V/P switchover point adjustment; solid line: real molding)

Full size image

After adjustment to the metering stroke, injection speed, and V/P switchover point, the difference in metering stroke was 16.3% (real molding: 52.97 mm; simulation: 44.3 mm). The difference in the foremost position of the screw was 13.3% (real molding: 6.18 mm; simulation: 5.36 mm). The difference in peak system pressure was 4.9% (real molding: 163 MPa; simulation: 171 MPa). The difference in peak P3 pressure was 5.7% (real molding: 70 MPa; simulation: 74 MPa). The difference in peak P8 pressure was 6.3% (real molding: 32 MPa; simulation: 30 MPa).

5.3 Width quality

Table 4 lists the width quality of the injection-molded part predicted by the simulation and that observed in the real molding. Figure 14 presents a comparison of their width shrinkage rates.

  1. 1.

    Under identical process parameter settings, the system pressure in the simulation was higher than that of the real molding, resulting in a larger width in the simulation. The negative value for the shrinkage rate in the simulation indicated an increase in width caused by overpacking.

  2. 2.

    After adjustment to the metering stroke and injection speed, the consistency between the simulated and real molding system pressures led to similar width quality.

  3. 3.

    After adjustment to the V/P switchover point, the consistency between the simulated and real molding cavity pressure was expected to produce similar width quality. However, the real cavity pressure was lower than the simulated value, resulting in a high shrinkage rate. Adjusting the V/P switchover point can affect the holding conditions in the simulation and result in a low shrinkage rate. The improvements in W3 and W5 were notable. The geometrical locations of W1 and W4 may have prevented the transfer of holding pressure, resulting in ineffective packing.

Table 4 Width quality
Full size table
Fig. 14
figure 14

Comparison of width shrinkage rate between real molding and simulation with adjusted parameters

Full size image

6 Conclusion

The accuracy of mold flow simulation is essential to intelligent injection mold manufacturing. However, simulation technology is limited by several factors, resulting in inconsistencies when compared with real molding. This study applied in-mold machine sensing information to determine the optimal process parameter settings for real molding and then adjusted the process parameters in the simulation on the basis of the real pressure and screw position curves, thereby decreasing the difference between the simulated and real pressure curves. The results are as follows:

  1. 1.

    Before the process parameter adjustment, the difference in the foremost position of the screw was 83% (real molding: 6 mm; simulation: 11 mm). The difference in peak system pressure was 47% (real molding: 163 MPa; simulation: 240 MPa). The system pressure in the simulation was considerably higher than that in the real molding, resulting in a larger simulated width. The negative value for the shrinkage rate in the simulation indicated an increase in width caused by overpacking.

  2. 2.

    After adjusting the metering stroke and injection speed, the difference in the metering stroke decreased to 16.3% (real molding: 52.97 mm; simulation: 44.3 mm). The difference in the foremost position of the screw was only 0.8% (real molding: 6.18 mm; simulation: 6.13 mm). The difference in peak system pressure decreased to 4.9% (real molding: 163 MPa; simulation: 171 MPa). The difference in peak P3 pressure was only 1.4% (real molding: 70 MPa; simulation: 71 MPa). However, the difference in peak P8 pressure was 93.8% (real molding: 32 MPa; simulation: 2 MPa). Overall, except for the far-from-gate cavity pressure curves, the consistency between the simulation and real molding system pressure led to similar width quality.

  3. 3.

    After adjusting the V/P switchover point, the difference in peak P8 pressure decreased considerably to 6.3% (real molding: 32 MPa; simulation: 30 MPa). The differences in metering stroke, foremost position of the screw, peak system pressure, and peak P3 pressure increased to 16.3%, 13.3%, 4.9%, and 5.7%, respectively. The real cavity pressure was lower than the simulated value, resulting in a high shrinkage rate. Adjusting the V/P switchover point can affect the holding conditions in the simulation and result in a low shrinkage rate. The improvements in W3 and W5 were notable. The geometrical locations of W1 and W4 may have prevented the transfer of holding pressure, resulting in ineffective packing.

  4. 4.

    This study demonstrated the potential of the proposed method to enable reduction of the difference between the simulated and real pressure and screw position curves through adjustment of the process parameter settings during a simulation. Subsequent studies should develop a fast and systematic method to identify the optimal process parameter settings during the simulation of various injection molding applications.