Control unit and thermal printer

Abstract

To provide a control unit which improves convenience by estimating the service life of a thermal print head.SOLUTION: A control unit (11) used for a thermal printer (100) comprising a thermal print head (1) which has a heating part (31) and is so constituted as to print by making a thermal reactor (8) react on heating of the heating part, comprises: a machine learning model (11A); and a machine learning part (11B) which performs teacher learning by inputting a print condition for the machine learning model, and by outputting a parameter related on the service life of the thermal print head for the machine learning model, and is so constituted as to infer the parameter on the basis of a learning result.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present disclosure relates to a control device. [Background technology]

[0002] Thermal printers are well known. Thermal printers are equipped with a thermal printhead. A thermal printhead is a device that prints by causing a thermally reactive material, such as thermal paper or a thermal transfer ribbon, to react with Joule heat generated by passing current through a resistor on a substrate (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-114710

[0004] [overview] Currently, thermal printhead maintenance involves replacing the thermal printhead periodically or after it breaks down. However, this method has problems such as premature replacement of the thermal printhead or sudden failure that makes the thermal printhead unusable.

[0005] In view of the above circumstances, an object of the present disclosure is to provide a control device that improves convenience by estimating the lifespan of a thermal printhead.

[0006] A control device according to one aspect of the present disclosure includes: A control device used in a thermal printer having a thermal print head configured to perform printing by reacting a thermal reactant with a heat generating unit, the control device comprising: Machine learning models and a machine learning unit configured to perform supervised learning using printing conditions as inputs to the machine learning model and parameters related to the life of the thermal print head as outputs of the machine learning model, and to infer the parameters based on the learning results; The configuration is provided with the following. [Brief explanation of the drawings]

[0007] [Figure 1] FIG. 1 is a diagram illustrating a thermal printhead according to an exemplary embodiment of the present disclosure. [Figure 2] FIG. 2 is a diagram showing a schematic diagram of the relationship between the rate of change in resistance value of the heat generating portion and the cumulative applied energy. [Figure 3] FIG. 3 is a block diagram of a thermal printer according to an exemplary embodiment of the present disclosure. [Figure 4] FIG. 4 is a diagram illustrating an example of the configuration of the voltage measurement unit. [Figure 5] FIG. 5 is a diagram illustrating another example of the configuration of the voltage measurement unit. [Figure 6] FIG. 6 is a diagram illustrating input and output to a machine learning model. [Figure 7] FIG. 7 is a diagram illustrating an example of input data. [Figure 8] FIG. 8 is a diagram showing an example of printing on a print medium. [Figure 9] FIG. 9 is a diagram illustrating an example of output data. [Figure 10] FIG. 10 is a flowchart illustrating an example of the learning and inference process. [Figure 11] FIG. 11 is a diagram illustrating an example of the process shown in FIG. 10 being executed. [Figure 12] FIG. 12 is a flowchart illustrating another example of the learning and inference process. [Figure 13] FIG. 13 is a diagram illustrating an example of the process shown in FIG. 12 being executed. [Figure 14] FIG. 14 is a diagram illustrating a three-layer neural network. [Figure 15]FIG. 15 is a graph showing an example of the learning and inference process. [Figure 16] FIG. 16 is a block diagram of a thermal printer according to a modified example.

[0008] [Detailed explanation] Hereinafter, exemplary embodiments of the present disclosure will be described with reference to the drawings.

[0009] <Thermal printer configuration> 1 shows a thermal printhead 1 according to an exemplary embodiment of the present disclosure. The thermal printhead 1 of this embodiment includes a substrate 2, a glaze layer (not shown), a wiring layer (not shown), a resistor layer 3, a protective layer (not shown), a driving IC 4, a protective resin 5, and a connector 6.

[0010] In FIG. 1, the thickness direction of the substrate 2 is referred to as the "thickness direction z." The top of FIG. 1 is the "one side z1 in the thickness direction," and the bottom of FIG. 1 is the "other side z2 in the thickness direction." The main scanning direction of the thermal printhead 1 is referred to as the "main scanning direction x," and the sub-scanning direction of the thermal printhead 1 is referred to as the "sub-scanning direction y." With respect to the sub-scanning direction y, the left side of FIG. 1 is the upstream side where the print medium 8 is fed, and is referred to as the "one side y1 in the sub-scanning direction." The right side of FIG. 1 is the downstream side where the print medium 8 is discharged, and is referred to as the "other side y2 in the sub-scanning direction."

[0011] The thermal printhead 1 is provided in a thermal printer 100 that prints on a print medium 8. The thermal printer 100 includes the thermal printhead 1 and a platen roller 7. The platen roller 7 faces the thermal printhead 1. The print medium 8 is sandwiched between the thermal printhead 1 and the platen roller 7 and is transported in the sub-scanning direction y by the platen roller 7. The print medium 8 may be, for example, thermal paper for creating barcode labels or receipts. Alternatively, a transfer paper and ribbon may be fed between the thermal printhead and the platen roller, and ink applied to the ribbon may be transferred to the transfer paper by heat.

[0012] The substrate 2 is made of ceramic. The glaze layer, wiring layer, resistor layer 3, protective layer, drive IC 4, and protective resin 5 are each disposed on the substrate 2. The connector 6 is for connecting to external devices and is provided at one end of the substrate 2 in the sub-scanning direction.

[0013] The glaze layer is disposed on the substrate 2 and is made of a glass material. The thermal printhead 1 has a so-called thick-film configuration and is manufactured using thick-film printing. The glaze layer is formed by printing a thick film of glass paste on the substrate 2 and then firing it. The glaze layer 2 is formed using thick-film formation technology.

[0014] The wiring layer is disposed on the glaze layer and serves to form a path for passing current through the resistor layer 3. The wiring layer is formed to have a resistivity smaller than that of the resistor layer 3. The wiring layer is made of a conductor containing, for example, Ag (silver) as a main component.

[0015] The resistor layer 3 is made of a material with a higher resistivity than the material constituting the wiring layer, such as ruthenium oxide, and is formed in a strip shape extending in the main scanning direction x. In a plan view seen in the thickness direction z, the resistor layer 3 has a plurality of heat generating portions 31 arranged along the main scanning direction x. More specifically, the wiring layer has a common electrode portion and a plurality of individual electrode portions. Each heat generating portion 31 is connected to the common electrode portion and each individual electrode portion. The plurality of heat generating portions 31 are selectively heated by being partially energized by the driving IC 4. In other words, one print dot is formed by the heat generated by one heat generating portion 31.

[0016] The protective layer is intended to protect at least the resistor layer 3, and contains glass such as amorphous glass as a main component. The driver IC 4 is covered with a protective resin 5.

[0017] The thermal printhead 1 is used in a thermal printer 100. As shown in FIG. 1, each heat-generating element 31 of the thermal printhead 1 faces a platen roller 7. When the thermal printer 100 is in use, the platen roller 7 rotates, feeding the print medium 8 at a constant speed between the platen roller 7 and each heat-generating element 31 along the sub-scanning direction y. The platen roller 7 presses the print medium 8 against the portions of the protective layer that cover each heat-generating element 31. Meanwhile, a potential is selectively applied to each individual electrode by the drive IC 4. This applies a voltage between the common electrode and each individual electrode. Current then flows selectively through the multiple heat-generating elements 31, generating heat. The heat generated by the heat-generating elements 31 is then transferred to the print medium 8 via the protective layer. Multiple dots are then printed in a linear region on the print medium 8 that extends linearly in the main scanning direction x. The heat generated by each heat-generating element 31 is also transferred to the glaze layer and stored there.

[0018] <About the lifespan of a thermal print head> The lifespan of a thermal printhead can be estimated based on changes in the resistance of the heat-generating element in the thermal printhead. For example, when the rate of change in resistance calculated based on the resistance R of the heat-generating element and the initial value Rini of that resistance R as (R-Rini) / Rini x 100% reaches a predetermined ratio (e.g., 15%) or more, it can be estimated that the thermal printhead has reached the end of its lifespan. In particular, the resistance of the heat-generating element tends to change slowly in thick-film thermal printheads, making it easy to predict their lifespan.

[0019] Figure 2 is a diagram showing the relationship between the rate of change in resistance value of the heat generating element and the cumulative applied energy. Note that the applied energy refers to the energy (product of power and time) applied to raise the temperature of the heat generating element to the temperature required for printing. As shown in Figure 2, the rate of change in resistance value increases as the cumulative applied energy increases.

[0020] 2 shows the relationship between the resistance change rate and the cumulative applied energy for each printing cycle condition. The printing cycle (SLT (Scanning Line Time)) refers to the cycle for driving the heat generating element and is expressed by the following formula: SLT=1 / (DPI×IPS) Here, DPI (Dots / inch) is the dot density, expressed in the number of dots per inch, and IPS (Inch Per Second) is the printing speed, expressed in inches of distance printed per second.

[0021] In Figure 2, the solid line indicates a large printing cycle, the dashed line indicates a medium printing cycle, and the thin dashed line indicates a small printing cycle. As can be seen, even with the same cumulative applied energy, the smaller the printing cycle, the greater the rate of change in resistance value.

[0022] The resistance change rate varies depending on printing conditions such as the printing cycle, applied energy, and print pattern (described later). In reality, the thermal printhead is used with different printing conditions for each printing operation (for example, printing on a single receipt), so it is not easy to estimate the resistance change rate.

[0023] Therefore, in this embodiment, the rate of change in the resistance value of the heating element is estimated by performing machine learning and inference using various printing conditions, including measurements of the resistance value of the heating element, and a machine learning model (AI model). This makes it possible to estimate the lifespan of the thermal printhead with high accuracy. This embodiment will be described in more detail below.

[0024] <Block configuration> Figure 3 is a diagram showing a schematic block configuration of a thermal printer 100 according to an embodiment of the present disclosure. In addition to the thermal printhead 1 described above, the thermal printer 100 shown in Figure 3 also includes an MCU (microcontroller) 11, a system controller 12, a voltage measurement unit 13, and a warning unit 14. Note that the block diagrams that follow, including Figure 3, mainly illustrate functional units related to estimating the lifespan of the thermal printhead, and omit, for example, the driver for the motor that drives the platen roller 7. Similarly, for the internal configuration of the MCU 11, some functional units are omitted for convenience.

[0025] The MCU 11 is a control device that performs overall control of the thermal printer 100. The system controller 12 is configured as an ASIC (application specific integrated circuit), and controls the thermal printhead 1 (drive control of the heat generating portion) based on commands from the MCU 11.

[0026] The MCU 11 includes a machine learning model 11A, a machine learning unit 11B, a warning control unit 11C, and a resistance value detection unit 11D.

[0027] The machine learning model 11A is configured by a neural network as described below. The machine learning unit 11B performs learning using the machine learning model 11A and printing conditions as described below. The machine learning unit 11B also infers the rate of change in the resistance value of the heat generating portion based on the learning results.

[0028] The warning control unit 11C performs warning control using the warning unit 14 based on the inferred rate of change in resistance value. If the warning unit 14 is configured with, for example, a display unit (e.g., an LCD (liquid crystal display)), it warns the user by display. If the warning unit 14 is configured with, for example, an audio output unit (e.g., a speaker), it warns the user by audio. Specific examples of warning control will be described later.

[0029] <<Resistance measurement>> In this embodiment, the measured value of the resistance value of the heat generating portion 31 is used for learning by the machine learning model 11A. Therefore, in the configuration shown in Fig. 3, a voltage measuring unit 13 and a resistance value detecting unit 11D are provided.

[0030] 4 is a diagram showing an example of the configuration of the voltage measurement unit 13. Here, the voltage measurement unit 13 has a constant current source 13A for supplying a constant current Ic to the heat generation unit 31. This generates a voltage Vdet across the heat generation unit 31. The resistance value detection unit 11D detects the resistance value of the heat generation unit as resistance value R [Ω] = Vdet / Ic.

[0031] 5 is a diagram showing another example of the configuration of the voltage measurement unit 13. Here, the voltage measurement unit 13 has a shunt resistor 13B connected in series to the heat generation unit 31. In FIG. 5, the shunt resistor 13B is connected between the heat generation unit 31 and a ground terminal (a terminal to which the ground potential is applied). An applied voltage Vcc is applied to the heat generation unit 31. This causes a current to flow through the shunt resistor 13B, and a voltage Vis is generated across the shunt resistor 13B. The voltage Vis output from the voltage measurement unit 13 is sent to the resistance value detection unit 11D.

[0032] The resistance value detection unit 11D detects the resistance value of the heat generating unit 31 as follows. The current Ia flowing through the heat generating unit 31 is Ia [A] = Vis / Rst, where Rst is the resistance value of the shunt resistor 13B. Then, if R is the resistance value of the heat generating unit 31, Vcc - Vis = R x Ia, and therefore R [Ω] = (Vcc - Vis) / Ia. Note that the shunt resistor is not limited to the configuration shown in Figure 5, and may be connected between the application terminal of the applied voltage Vcc and the heat generating unit 31.

[0033] <Temperature measurement> As shown in Figure 3, the thermal printer 100 also includes a temperature measurement unit 15. The temperature measurement unit 15 has a thermistor and measures the temperature of the substrate 2 in the thermal printhead 1 based on the change in resistance value of the thermistor due to temperature. The temperature measurement result is sent to the MCU 11, which then monitors the temperature of the substrate 2.

[0034] <About the data> 6, input data x is input to the machine learning model 11A, and output data y is output from the machine learning model 11A. During inference, various printing conditions are used as the input data x, and the rate of change in resistance value of the heat generating part 31 is output as the output data y. During learning, various printing conditions are used as the input data x, the output data y is set as training data t, and the rate of change in resistance value is used as the training data t.

[0035] Fig. 7 is a diagram showing an example of data of printing conditions used as input data x. As shown in Fig. 7, the input data x includes the measured resistance value of the heat generating element 31, the measured temperature of the substrate 2, the printing cycle, the applied energy, the number of prints, and the print pattern.

[0036] The number of prints refers to the number of lines, where one print is one line of dots formed by multiple heating elements 31 arranged along the main scanning direction x. FIG. 8 is a diagram showing an example of printing on a print medium 8. FIG. 8 shows one line L extending along the main scanning direction x. In FIG. 7, data obtained when printing is performed a predetermined number of times (here, 10 times as an example) is used as input data x.

[0037] The measured resistance value is the resistance value detected by the resistance value detection unit 11D. The measured temperature value is the temperature measured by the temperature measurement unit 15. Note that the resistance value or temperature may be, for example, the average value of multiple measured values. As mentioned above, the printing cycle indicates the drive cycle of the heat generating unit 31. The applied energy may be, for example, the accumulated energy in all heat generating units 31 when printing is performed, or the accumulated energy in only the heat generating unit 31 whose resistance value is to be measured.

[0038] The print pattern is data that indicates whether or not each dot is printed when printing is performed. In the example of Fig. 8, printed dots are indicated by black circles, and unprinted dots are indicated by white circles. The print pattern may be, for example, the ratio of the number of printed (or unprinted) dots to the total number of dots when printing is performed.

[0039] FIG. 9 is a diagram showing data used for output data y. The output data y indicates time-series data of the resistance change rate of the heat generating element 31 after the printing operation corresponding to the input data x is executed. In other words, when the input data x is input to the machine learning model 11A, the future behavior of the resistance change rate is inferred. The output data y is, for example, data indicating the resistance change rate for a predetermined number of prints (e.g., 10 prints). As shown in FIG. 7, for input data x for 100 prints, the output data y can be data for the subsequent 500 prints, for example.

[0040] During learning, training data t is used instead of output data y, and training data t is the same as that shown in Figure 9. However, in this case, the resistance change rate is calculated based on the measured resistance value (actual value).

[0041] <Learning and inference methods> An example of the learning and inference processing in this embodiment will now be described. Note that the specific values ​​of the number of prints described here are merely examples. FIG. 10 is a flowchart relating to an example of the learning and inference processing. A first form of processing will be described along FIG. 10 with reference to FIG. 11 as well. Note that FIG. 11 shows the progress of processing vertically, and the number of prints on the horizontal axis.

[0042] When the process starts, input data x for 100 prints is acquired (step S1). Next, output data y is inferred based on the acquired input data x and machine learning model 11A (step S2). Here, as shown in (1) of FIG. 11, input data INDATA1 for 100 prints is acquired, and output data OUTDATA1 for the next 500 prints is inferred.

[0043] Then, training data (resistance change rate) and input data x for the next 100 print runs are acquired (step S3). Next, learning is performed using the input data x for the 100 print runs acquired immediately before the most recent one and training data t obtained by replacing the first 100 of the data inferred above with the training data (step S4). Here, as shown in (2) of Figure 11, learning is performed using training data t (500 runs) obtained by replacing the first 100 of the input data INDATA1 and OUTDATA1 with training data TRNDATA1.

[0044] Then, the process returns to step S2, and inference is performed based on the input data x (the most recent 100 pieces of data) acquired in step S3. Here, as shown in (3) of FIG. 11, output data OUTDATA2 is inferred based on input data INDATA2.

[0045] Then, in step S3, training data and input data x for the next 100 prints are acquired, and learning is performed in step S4. Here, as shown in (4) of Fig. 11, learning is performed using training data t obtained by replacing the input data INDATA2 and the first 100 prints of OUTDATA2 with training data TRNDATA2.

[0046] Then, the process returns to step S2, and inference is performed based on the input data x (the most recent 100 pieces of data) acquired in step S3. Here, as shown in (5) of Fig. 11, output data OUTDATA3 is inferred based on input data INDATA3. Thereafter, the same operation is repeated.

[0047] In this way, the next 500 inferences (OUTDATA1, 2, 3) are performed based on the most recent 100 input data sets (INDATA1, 2, 3) acquired.

[0048] Fig. 12 is a flowchart showing another example of the learning and inference process. A second form of the process will be described with reference to Fig. 13 along with Fig. 12. Note that Fig. 13 shows the progress of the process vertically, and the horizontal axis shows the number of prints.

[0049] When the process starts, input data x for 100 prints is acquired (step S11). Next, output data y is inferred based on the acquired input data x and machine learning model 11A (step S12). Here, as shown in FIG. 13 (1), input data INDATA1 for 100 prints is acquired, and output data OUTDATA1 for the next 500 prints is inferred.

[0050] Then, training data (resistance change rate) for the next 100 prints is acquired (step S13). Next, learning is performed using the acquired input data x and training data t obtained by replacing the first 100 prints of the data inferred above with the training data (step S14). Here, training data is added at each step and replaced with the accumulated training data. Here, as shown in (2) of FIG. 11, learning is performed using training data t (500 prints) obtained by replacing the first 100 prints of input data INDATA1 and OUTDATA1 with training data TRNDATA1.

[0051] Then, the process returns to step S12, and inference is performed based on the input data x acquired in step S11. Here, as shown in (3) of Fig. 13, output data OUTDATA2 is inferred based on input data INDATA1.

[0052] Then, in step S13, training data for the next 100 prints is acquired, and learning is performed in step S14. Here, as shown in (4) of Fig. 13, learning is performed using training data t in which the input data INDATA1 and the first 200 prints of OUTDATA2 are replaced with training data TRNDATA1 and TRNDATA2. In other words, TRNDATA2 is added to training data TRNDATA1 and replaced with the accumulated training data.

[0053] Then, the process returns to step S12, where inference is performed based on the input data x acquired in step S11. Here, as shown in (5) of Fig. 13, output data OUTDATA3 is inferred based on input data INDATA1. Thereafter, similar operations are repeated.

[0054] In this way, inference is performed based on the first 100 input data (INDATA1), and learning is performed while increasing the proportion of actual training data (TRNDATA1,2) in the 500 training data, so the accuracy of learning improves over time.

[0055] <Sequential learning processing method> Here, an example of the machine learning model 11A will be described. As an AI model used in the machine learning model 11A, for example, a three-layer neural network 50 as shown in FIG.

[0056] 14, a three-layer neural network 50 is an AI model having an input layer 50A, a hidden layer 50B, and an output layer 50C. In general, in the three-layer neural network 50, n-dimensional input data x∈R with a batch size k is input. k×n For n'-dimensional inference result y∈R k×n’ is obtained as y=G(x·α+b)β, where α∈R n×m is the weight connecting the input layer 50A and the hidden layer 50B, and β∈R m×n’ is the weight connecting the hidden layer 50B and the output layer 50C. mis the bias of hidden layer 50B, and G is the activation function of hidden layer 50B.

[0057] In this embodiment, an algorithm that can be sequentially learned using a three-layer neural network 50 is used. i The i-th training data {x i ∈R ki×n , t i ∈R ki×n’} is obtained, β that minimizes the error shown in the following equation (1) i It is necessary to seek.

number

[0058] Optimized weight β i is calculated using the following formula (2). P i =P i-1 -P i-1 H i T (I+H i P i-1 H i T ) -1 H i P i-1 β i =β i-1 +P i H i T (t i -H i β i-1 ) (2)

[0059] Here, P0 and β0 are obtained by the following equation (3). P0=(H0 T H0) -1 β0=P0H0 T t0(3)

[0060] The learning algorithm is as follows: (1) The weight α and bias b are initialized with random numbers. (2) Calculate H0 for x0, and calculate P0 and β0. (3) Batch size k i Each time the i-th training data of P is obtained, i and β i It should be noted that instead of using the formula for calculating β0 in equation (3), a value initialized by a random number may be used as β0.

[0061] Here, the computational bottleneck of equation (2) is the inverse matrix operation (I + H i P i-1 H i T ) -1 However, (I+H i P i-1 H i T Since the matrix size of (k×k) is k×k, if k=1, the inverse matrix operation can be replaced with the reciprocal operation. Therefore, in this embodiment, it is desirable to fix k=1. This makes it easier for the MCU 11 to perform the learning operation.

[0062] Furthermore, an RNN (Recurrent Neural Network) may be used for learning. Using an RNN makes it possible to learn trends in long-term data. In particular, in this embodiment, it is preferable to use an ESN (Echo State Network), which is a type of RNN that enables sequential learning.

[0063] The ESN consists of an input layer, a hidden layer, and an output layer. i is obtained, the output of the hidden layer in the ESN model is calculated by the following equation (4): i is calculated. h i =G((1-δ)h i-1 +δ(x i α+h i-1 γ)) (4) α is the weight connecting the input layer and the hidden layer, γ is the feedback weight connecting the hidden layers, δ is the leak rate, and G is the activation function.

[0064] Here, when RLS (Recursive Least Squares) is applied to ESN, the loss function is expressed by the following equation (5).

number

[0065] where:

number

[0066] p i ≡k i -1 and q i ≡(1 / λ)p i Then, the above equation (6) finally becomes the following equation (7): This enables sequential learning. p i =q i-1 -q i-1 h i T (I+h i q i-1 h i T )-1 h i q i-1 β i =β i-1 +q i h i T (t i -h i β i-1 ) (7)

[0067] β i is the previous intermediate result p i-1 Since it can be calculated from the above, there is no need for memory to store past learning data.

[0068] <About warning control> Fig. 15 is a graph showing an example of the first mode (Fig. 11) or the second mode (Fig. 13) of the learning and inference process described above. In Fig. 15, the horizontal axis represents the number of prints, and the vertical axis represents the rate of change in resistance value.

[0069] In the first mode, inference is performed for the next 500 prints based on the most recent 100 prints of input data (print conditions 1, 2, and 3). In the second mode, inference is performed based on the first 100 prints of input data (print condition 1), while learning is performed while increasing the proportion of actual data (print conditions 2 and 3) in the training data for 500 prints. In Figure 15, as an example, the resistance change rate, which indicates the lifespan, is set to a lifespan value Rx slightly lower than 20%, and the inferred resistance change rate (dashed line) reaches the lifespan value Rx after printing count Px.

[0070] When the number of prints (life print count) at which the resistance change rate reaches the life value is determined based on the inference result by the machine learning unit 11B, the warning control unit 11C performs the following control. The above-mentioned number of prints Px is an example of the life print count. For example, if the current number of prints is equal to or greater than a predetermined value that is a predetermined number lower than the life print count, the warning control unit 11C causes the warning unit 14 to warn that the life will soon be reached. This allows the user to perform repairs in advance. Therefore, since maintenance is performed when the product is nearing the end of its original life, the product life can be extended as a result. Furthermore, a situation in which the product suddenly becomes unusable can also be avoided. At this time, the number of prints from the current number of prints to the life print count may be notified.

[0071] <Modification> Fig. 16 is a block diagram showing the configuration of a modified thermal printer 100. The configuration shown in Fig. 16 differs from the configuration shown in Fig. 3 in that the MCU 11 is provided with an energy consumption detection unit 11E instead of the resistance value detection unit 11D.

[0072] The voltage measurement unit 13 outputs the voltage Vdet of the heat generating unit 31 as in the configuration shown in Fig. 4. The energy consumption detection unit 11E detects the energy consumption E as E[J] = Vdet × Ic × t, where Ic is the constant current Ic shown in Fig. 4, and t is the time during which the voltage is applied to the heat generating unit 31.

[0073] The energy consumption measured in this way is used in place of the resistance value in input data such as that shown in FIG. 7. The energy consumption is a parameter that can indirectly estimate the resistance value. The same applies to the temperature (heat generation amount) shown in FIG. 7. The input data can include at least one of the resistance value, the energy consumption, and the temperature.

[0074] <Other> In addition to the above-described embodiments, the various technical features disclosed in this specification can be modified in various ways without departing from the spirit of the technical creation. In other words, the above-described embodiments should be considered to be illustrative and not restrictive in all respects, and the technical scope of the present invention should not be limited to the above-described embodiments, but should be understood to include all modifications that fall within the meaning and scope equivalent to the claims.

[0075] <Additional Notes> As described above, the control device (11) according to one aspect of the present disclosure includes: A control device used in a thermal printer (100) having a thermal printhead (1) configured to perform printing by having a heat generating unit (31) and causing a thermal reactant (8) to react when the heat generating unit generates heat, Machine learning model (11A) and a machine learning unit (11B) configured to perform supervised learning by using printing conditions as inputs to the machine learning model and parameters related to the life of the thermal print head as outputs of the machine learning model, and to infer the parameters based on the learning results; (first configuration).

[0076] According to this configuration, the parameters relating to the life of the thermal printhead can be estimated with high accuracy, thereby improving the convenience of the thermal printer.

[0077] In the first configuration, the print conditions may include a parameter relating to the resistance value of the heat generating portion (second configuration).

[0078] In the second configuration, the parameter relating to the resistance value may be the resistance value itself, temperature, or energy consumption (third configuration).

[0079] Furthermore, the third configuration may be configured to include a resistance value detector (11D) configured to measure the resistance value of the heat generating portion (fourth configuration).

[0080] Furthermore, the device may be configured to include an energy consumption detector (11E) configured to measure the energy consumption of the heat generating part (fifth configuration).

[0081] In any of the first to fifth configurations, the printing conditions may include at least one of a printing cycle, applied energy, the number of prints, and a print pattern (sixth configuration).

[0082] In any one of the first to sixth configurations, the parameter relating to the life of the thermal printhead may be a parameter based on the amount of change in resistance value of the heat generating element (seventh configuration).

[0083] In addition, in any one of the first to seventh configurations, the machine learning unit Inference processing based on input data for the latest predetermined number of prints; It may also be configured to execute a learning process in which a portion of the results of the inference process is replaced with the most recently acquired teacher data for the specified number of prints and the resulting data is used as teacher data for learning (eighth configuration).

[0084] In addition, in any one of the first to seventh configurations, the machine learning unit an inference process based on input data for the first predetermined number of prints; It may also be configured to execute a learning process in which a portion of the results of the inference process is replaced with accumulated teacher data while adding the most recently acquired teacher data for the specified number of prints, and the resulting data is used as teacher data for learning (ninth configuration).

[0085] Furthermore, in any of the first to ninth configurations, a configuration may be provided (tenth configuration) that includes a warning control unit (11C) configured to control warnings regarding the lifespan based on the inference results by the machine learning unit.

[0086] In addition, in any of the first to tenth configurations, the machine learning model may be configured as a three-layer neural network (50) having an input layer (50A), a hidden layer (50B), and an output layer (50C) (eleventh configuration).

[0087] In the eleventh configuration, the machine learning unit calculates the weight β connecting the hidden layer and the output layer using the following formula: i Alternatively, the learning may be performed sequentially by sequentially calculating (twelfth configuration). P i =P i-1 -P i-1 H i T (I+H i P i-1 H i T ) -1 H i P i-1 β i =β i-1 +P i H i T (t i -H i β i-1 ) However, the hidden layer matrix H i =G(x i ·α+b), α: weight connecting the input layer and the hidden layer, b: bias of the hidden layer, G: activation function of the hidden layer, x i : batch size k i The i-th input data, t i : ith training data of batch size ki

[0088] In the eleventh configuration, the machine learning model may be an ESN, which is a type of RNN (thirteenth configuration).

[0089] A thermal printer (100) according to an aspect of the present disclosure includes the control device of any one of the first to thirteenth configurations and the thermal printhead (fourteenth configuration). [Industrial Applicability]

[0090] The present disclosure can be used in thermal printers. [Explanation of symbols]

[0091] 1 Thermal print head 2 boards 3 Resistor layer 4 Driver IC 5 Protective resin 6 Connectors 7 Platen roller 8 Print media 11 MCU 11A Machine Learning Models 11B Machine Learning Department 11C Warning control section 11D Resistance detection unit 11E Energy consumption detector 12 System Controller 13 Voltage measurement section 13A constant current 13B Shunt Resistor 14 Warning part 15 Temperature measurement part 31 Heat generating part 50 Three-layer neural network 50A input layer 50B Hidden layer 50C output layer 100 Thermal Printer

Claims

1. A control device used in a thermal printer having a thermal print head configured to perform printing by reacting a thermal reactant with a heat generating unit, the control device comprising: Machine learning models and a machine learning unit configured to perform supervised learning using printing conditions as inputs to the machine learning model and parameters related to the life of the thermal print head as outputs of the machine learning model, and to infer the parameters based on the learning results; A control device comprising:

2. The control device according to claim 1 , wherein the printing conditions include a parameter relating to a resistance value of the heat generating portion.

3. The control device according to claim 2 , wherein the parameter relating to the resistance value is the resistance value itself, a temperature, or energy consumption.

4. The control device according to claim 3 , further comprising a resistance value detection unit configured to measure a resistance value of the heat generating portion.

5. The control device according to claim 3 , further comprising an energy consumption detection unit configured to measure the energy consumption of the heat generating unit.

6. 2. The control device according to claim 1, wherein the printing conditions include at least one of a printing cycle, applied energy, number of prints, and print pattern.

7. 2. The control device according to claim 1, wherein the parameter relating to the life of the thermal printhead is a parameter based on an amount of change in resistance value of the heating element.

8. The machine learning unit Inference processing based on input data for the latest predetermined number of prints; 2. The control device according to claim 1, further configured to execute a learning process in which a part of the result of the inference process is replaced with the most recently acquired teacher data for the predetermined number of prints, and the result is used as the teacher data for learning.

9. The machine learning unit an inference process based on input data for the first predetermined number of prints; The control device according to claim 1, configured to execute a learning process in which a portion of the results of the inference process is replaced with accumulated teacher data while adding the most recently acquired teacher data for the specified number of prints, and the resulting teacher data is used as teacher data for learning.

10. The control device according to claim 1 , further comprising a warning control unit configured to control a warning regarding the lifespan based on an inference result by the machine learning unit.

11. The control device of claim 1 , wherein the machine learning model is a three-layer neural network having an input layer, a hidden layer, and an output layer.

12. The machine learning unit calculates the weight β connecting the hidden layer and the output layer using the following formula: ｉ The control device according to claim 11 , wherein the control device performs sequential learning by sequentially calculating P ｉ =P ｉ－１ -P ｉ－１ H ｉ T (I+H ｉ P ｉ－１ H ｉ T ) －１ H ｉ P ｉ－１ b ｉ =b ｉ－１ +P ｉ H ｉ T (t) ｉ -H ｉ b ｉ－１ ) However, the hidden layer matrix H ｉ = G(x ｉ α+b), α: weight connecting the input layer and the hidden layer, b: bias of the hidden layer, G: activation function of the hidden layer, x ｉ : batch size k ｉ The i-th input data of t ｉ : i-th training data of batch size ki

13. The control device according to claim 11 , wherein the machine learning model is an ESN, which is a type of RNN.

14. A control device according to any one of claims 1 to 13; A thermal printer comprising the thermal printhead.

Images