TWI819318B - Machine monitoring device and method - Google Patents

Abstract

A machine monitoring device is provided, which includes a transceiver circuit, a memory and a processor. The transceiver circuit is configured to receive first self-test data and first measurement data; the memory is configured to store multiple instructions; and the processor is connected to the transceiver circuit and the memory, and is configured to load and execute the multiple instructions to: generate a prediction model related to the first self-test data and the first measurement data; the transceiver circuit receives second self-test data, and uses the prediction model to generate prediction data based on the second self-test data; and based on the prediction data and the numerical range to generate monitoring results, to monitor based on the monitoring results. In addition, a machine monitoring method also disclosed herein.

Description

Translated fromChinese

機台監控裝置以及方法Machine monitoring device and method

本發明有關於一種監控技術，且特別是有關於一種機台監控裝置以及方法。The present invention relates to a monitoring technology, and in particular to a machine monitoring device and method.

當工廠或廠房中的機台生產產品時，對產品或機台進行量測是產品品質把關的重要防線。然而，量測裝置有時候會發生異常狀態。此時，量測裝置之量測的精準度便隨之下降，且量測裝置之量測資料也失去參考價值。因此，要如何得知量測裝置發生異常狀態以造成量測資料失去參考價值是本領域技術人員急欲解決的問題。When machines in factories or factories produce products, measuring the products or machines is an important line of defense for product quality control. However, measurement devices sometimes experience abnormal conditions. At this time, the measurement accuracy of the measurement device decreases, and the measurement data of the measurement device also loses its reference value. Therefore, how to know that the measurement device is in an abnormal state, causing the measurement data to lose reference value, is a problem that those skilled in the art are eager to solve.

本發明提供一種機台監控裝置，其包括收發電路、記憶體以及處理器。收發電路用以接收第一自檢資料以及第一量測資料，其中第一自檢資料相關於量測裝置，其中第一量測資料相關於由至少一機台所製造的多個第一產品；記憶體用以儲存多個指令；以及處理器連接收發電路以及記憶體，並用以載入並執行多個指令以：產生與第一自檢資料以及第一量測資料相關的預測模型；經由收發電路接收第二自檢資料，並依據第二自檢資料以利用預測模型產生預測資料；以及依據預測資料以及數值範圍產生監控結果，以依據監控結果進行監控。The invention provides a machine monitoring device, which includes a transceiver circuit, a memory and a processor. The transceiver circuit is used to receive first self-test data and first measurement data, wherein the first self-test data is related to the measurement device, and the first measurement data is related to a plurality of first products manufactured by at least one machine; The memory is used to store multiple instructions; and the processor is connected to the transceiver circuit and the memory, and is used to load and execute multiple instructions to: generate a prediction model related to the first self-test data and the first measurement data; through the transceiver The circuit receives the second self-test data, uses the prediction model to generate prediction data based on the second self-test data, and generates monitoring results based on the prediction data and the numerical range, so as to perform monitoring based on the monitoring results.

本發明提供一種機台監控方法，其中所述方法包括：藉由收發電路接收第一自檢資料以及第一量測資料，其中第一自檢資料相關於量測裝置，其中第一量測資料相關於由至少一機台所製造的多個第一產品；藉由處理器對第一量測資料以及第一自檢資料執行機器學習演算法以產生預測模型；藉由處理器經由收發電路接收第二自檢資料，並依據第二自檢資料以利用預測模型產生預測資料；以及藉由處理器依據預測資料以及數值範圍產生監控結果，以依據監控結果進行監控。The present invention provides a machine monitoring method, wherein the method includes: receiving first self-test data and first measurement data through a transceiver circuit, wherein the first self-test data is related to the measurement device, and wherein the first measurement data Related to a plurality of first products manufactured by at least one machine; the processor executes a machine learning algorithm on the first measurement data and the first self-test data to generate a prediction model; the processor receives the third product through thetransceiver circuit 2. self-test data, and use the prediction model to generate prediction data based on the second self-test data; and use the processor to generate monitoring results based on the prediction data and the value range, so as to perform monitoring based on the monitoring results.

基於上述，本發明提供的機台監控裝置以及方法可預測機台在未來所製造的產品的量測資料，並可即時地檢測量測裝置是否發生異常狀態。Based on the above, the machine monitoring device and method provided by the present invention can predict the measurement data of products manufactured by the machine in the future, and can instantly detect whether an abnormal state of the measuring device occurs.

第1圖是根據本發明一些示範性實施例的機台監控裝置100的方塊圖。參照第1圖，機台監控裝置100可以是任意的邊緣運算（Edge Computing）裝置（例如，智慧型手機、平板電腦、筆記型電腦、桌上型電腦或伺服器等電子裝置）。在實際應用中，機台監控裝置100可連接量測裝置200，量測裝置200可連接至少一機台300(1)~300(N)，其中N可以為任意的正整數。Figure 1 is a block diagram of amachine monitoring device 100 according to some exemplary embodiments of the present invention. Referring to FIG. 1 , themachine monitoring device 100 can be any edge computing device (for example, an electronic device such as a smart phone, a tablet, a notebook computer, a desktop computer, or a server). In practical applications, themachine monitoring device 100 can be connected to themeasuring device 200, and themeasuring device 200 can be connected to at least one machine 300(1)~300(N), where N can be any positive integer.

進一步而言，量測裝置200可對機台300(1)~300(N)所製造的多個產品進行量測以產生量測資料。量測裝置200也可進行自我檢測以產生自檢資料。如此一來，機台監控裝置100可從量測裝置200接收上述量測資料以及自檢資料。Furthermore, themeasuring device 200 can measure multiple products manufactured by the machines 300(1)~300(N) to generate measurement data. Themeasuring device 200 can also perform self-testing to generate self-testing data. In this way, themachine monitoring device 100 can receive the above-mentioned measurement data and self-test data from themeasurement device 200 .

舉例而言，量測裝置200可對上述產品的各輸出埠或輸入埠等進行量測以產生量測資料（例如，電壓、電流或功率等），也可對自身的各輸出埠或輸入埠等進行自我檢測以產生自檢資料（例如，電壓、電流或功率等），進而將此量測資料以及此自檢資料傳送至機台監控裝置100。For example, themeasuring device 200 can measure each output port or input port of the above-mentioned products to generate measurement data (for example, voltage, current or power, etc.), or it can also measure each output port or input port of itself. etc. to perform self-test to generate self-test data (for example, voltage, current or power, etc.), and then transmit the measurement data and the self-test data to themachine monitoring device 100 .

在一些實施例中，量測裝置200可在多個自檢時間進行自我檢測（即，週期性地自我檢測）。此外，在這些自檢時間之間的多個量測時間段，量測裝置200可對機台300(1)~300(N)所製造的多個產品進行量測。In some embodiments, themeasurement device 200 can perform self-testing at multiple self-testing times (ie, perform self-testing periodically). In addition, during multiple measurement time periods between these self-test times, themeasurement device 200 can measure multiple products manufactured by the machines 300(1)˜300(N).

舉例而言，第2圖是根據本發明一些實施例繪示自檢資料以及量測資料的示意圖。同時參照第1圖以及第2圖，量測裝置200可在自檢時間t執行自我檢測n（即，執行第n次自我檢測）以產生自檢資料n，並可在自檢時間t+1執行自我檢測n+1（即，執行第n+1次自我檢測）以產生自檢資料n+1。For example, FIG. 2 is a schematic diagram illustrating self-test data and measurement data according to some embodiments of the present invention. Referring to Figures 1 and 2 at the same time, themeasuring device 200 can perform self-test n (ie, perform the nth self-test) at self-test time t to generate self-test data n, and can perform self-test n at self-test time t+1. Perform self-test n+1 (ie, perform the n+1th self-test) to generate self-test data n+1.

此外，量測裝置200可在自檢時間t以及自檢時間t+1之間（即，上述量測時間段）執行產品量測k（即，執行第k次產品量測）至產品量測k+m（即，執行第k+m次產品量測）以產生量測資料k至量測資料k+m。換言之，在自檢時間t以及自檢時間t+1之間，量測裝置200可對機台300(1)~300(N)所生產的m+1個產品進行量測。In addition, themeasurement device 200 can perform product measurement k (ie, perform the k-th product measurement) between the self-test time t and the self-test time t+1 (ie, the above-mentioned measurement time period) to product measurement k+m (ie, perform k+m-th product measurement) to generate measurement data k to measurement data k+m. In other words, between the self-test time t and the self-test time t+1, themeasurement device 200 can measure m+1 products produced by the machines 300(1)~300(N).

以此類推，量測裝置200也可在其他自檢時間執行自我檢測，並可於其他相鄰兩個自檢時間之間進行產品量測。By analogy, themeasuring device 200 can also perform self-testing at other self-testing times, and can perform product measurement between other two adjacent self-testing times.

值得注意的是，上述量測資料以及自檢資料雖是藉由量測裝置200產生的，然而，在其他實施例中，上述量測資料以及自檢資料也可以不是藉由量測裝置200產生的，而是藉由機台300(1)~300(N)直接產生的（此時，機台300(1)~300(N)可視為上述的量測裝置200，且自檢資料可以是機台300(1)~300(N)對自身的各輸出埠或輸入埠等進行自我檢測以產生的）。It is worth noting that although the above measurement data and self-test data are generated by themeasurement device 200, in other embodiments, the above measurement data and self-test data may not be generated by themeasurement device 200. , but directly generated by the machines 300(1)~300(N) (at this time, the machines 300(1)~300(N) can be regarded as the above-mentionedmeasuring device 200, and the self-test data can be The machines 300(1)~300(N) perform self-detection on each of their own output ports or input ports to generate).

在本實施例中，機台監控裝置100可包括收發電路110、記憶體120以及處理器130。收發電路110可接收上述自檢資料以及量測資料。記憶體120可儲存多個指令。處理器130可連接收發電路110以及記憶體120，並用以載入並執行這些指令。In this embodiment, themachine monitoring device 100 may include atransceiver circuit 110, amemory 120, and aprocessor 130. Thetransceiver circuit 110 can receive the above-mentioned self-test data and measurement data. Thememory 120 can store multiple instructions. Theprocessor 130 can be connected to thetransceiver circuit 110 and thememory 120 and used to load and execute these instructions.

在一些實施例中，收發電路110例如是傳送器電路、類比-數位轉換器、數位-類比轉換器、低噪音放大器、混頻器、濾波器、阻抗匹配器、傳輸線、功率放大器、一個或多個天線電路以及本地儲存媒體元件的其中之一或其組合。In some embodiments, thetransceiver circuit 110 is, for example, a transmitter circuit, an analog-to-digital converter, a digital-to-analog converter, a low noise amplifier, a mixer, a filter, an impedance matcher, a transmission line, a power amplifier, one or more One or a combination of an antenna circuit and a local storage media element.

在一些實施例中，記憶體120可例如是任何型態的固定式或可移動式的記憶體、硬碟或類似元件或上述元件的組合。In some embodiments, thememory 120 may be, for example, any type of fixed or removable memory, a hard disk or similar components, or a combination of the above components.

在一些實施例中，處理器130例如是中央處理單元（Central Processing Unit，CPU），或是其他可程式化之一般用途或特殊用途的微控制單元（Micro Control Unit，MCU）、微處理器（Microprocessor）、數位信號處理器（Digital Signal Processor，DSP）、可程式化控制器、特殊應用積體電路（Application Specific Integrated Circuit，ASIC）或其他類似元件或上述元件的組合。In some embodiments, theprocessor 130 is, for example, a Central Processing Unit (CPU), or other programmable general-purpose or special-purpose Micro Control Unit (MCU), microprocessor ( Microprocessor), Digital Signal Processor (DSP), programmable controller, Application Specific Integrated Circuit (ASIC) or other similar components or a combination of the above components.

在一些實施例中，處理器130可以有線或無線的方式連接收發電路110與記憶體120。In some embodiments, theprocessor 130 may connect thetransceiver circuit 110 and thememory 120 in a wired or wireless manner.

第3圖是根據本發明一些示範性實施例的機台監控方法的示意圖。第3圖所示實施例的方法適用於第1圖的機台監控裝置100，但不以此為限。為方便及清楚說明起見，下述同時參照第1圖以及第3圖，以機台監控裝置100中各元件之間的作動關係來說明第3圖所示機台監控方法的詳細步驟。Figure 3 is a schematic diagram of a machine monitoring method according to some exemplary embodiments of the present invention. The method of the embodiment shown in Figure 3 is applicable to themachine monitoring device 100 in Figure 1 , but is not limited thereto. For the sake of convenience and clarity of explanation, the detailed steps of the machine monitoring method shown in Figure 3 will be described below with reference to Figures 1 and 3 at the same time, using the operational relationships between the components in themachine monitoring device 100.

首先，於步驟S310中，收發電路110可接收第一自檢資料以及第一量測資料，以儲存於記憶體120中的資料庫121。詳細而言，收發電路110可從量測裝置200接收量測裝置200所產生的第一自檢資料以及第一量測資料（即，量測裝置200過去所檢測到的自檢資料以及量測資料），以儲存於記憶體120中的資料庫121中。First, in step S310 , thetransceiver circuit 110 may receive the first self-test data and the first measurement data to be stored in thedatabase 121 in thememory 120 . In detail, thetransceiver circuit 110 can receive the first self-test data and the first measurement data generated by themeasurement device 200 from the measurement device 200 (ie, the self-test data and measurement data detected by themeasurement device 200 in the past). data) to be stored in thedatabase 121 in thememory 120.

在一些實施例中，第一自檢資料可對應於量測裝置200的第一自檢時間，且第一量測資料可對應於量測裝置200所量測之多個第一產品的第一量測時間段，其中第一量測時間段在第一自檢時間以及第二自檢時間之間。In some embodiments, the first self-test data may correspond to the first self-test time of themeasurement device 200 , and the first measurement data may correspond to the first time of the plurality of first products measured by themeasurement device 200 . A measurement time period, wherein the first measurement time period is between the first self-test time and the second self-test time.

換言之，量測裝置200可在第一自檢時間進行自我檢測以產生第一自檢資料。此外，在第一自檢時間以及第二自檢時間（即，在第一自檢時間之後的下一個自檢時間）之間的時間段，量測裝置200可量測機台300(1)~300(N)在此第一量測時間段所製造的多個第一產品以產生第一量測資料。In other words, themeasurement device 200 can perform self-test at the first self-test time to generate the first self-test data. In addition, during the time period between the first self-test time and the second self-test time (ie, the next self-test time after the first self-test time), themeasuring device 200 can measure the machine 300(1) ~300(N) multiple first products manufactured during the first measurement period to generate the first measurement data.

值得注意的是，本實施例雖僅示出與一個第一自檢時間對應的第一自檢資料以及與一個第一量測時間段對應的第一量測資料，然而，在實際應用上，也可以是與多個第一自檢時間對應的多個第一自檢資料以及與多個第一量測時間段對應的多個第一量測資料，並沒有特別的限制。舉例而言，以下表一示出了與多個第一自檢時間對應的多個第一自檢資料的例子，且以下表二示出了與多個第一量測時間段對應的多個第一量測資料的例子。 表一自檢編號第一自檢時間直流電壓V1直流電壓V2交流電壓V3直流電壓V4直流電壓V5直流電壓V6直流電流I1直流電壓V7直流電壓V8直流電壓V9直流電流I2接地電壓V002020/2/15 08:0112.00712.006220.010200.040200.07200.042.006350.043350.05349.981.0040.00312020/2/17 07:5212.01212.009220.022200.057200.08200.052.005350.086350.07350.001.0040.00322020/2/18 07:4512.01112.008220.021200.067200.04200.022.005350.084350.06349.981.0030.003. . .      . . .      表二第一量測時間產品序號高電壓的電壓（HV Voltage）高電壓的電流（HV current）輸入功率（P in）低電壓的電壓（LV Voltage）低電壓的電流（LV Current）輸出功率（P out）效率（Efficient）雜訊（Noise）2020/2/15 08:49A0001360.020.486175.1113.49910.992148.40584.7490.0312020/2/15 09:11A0002360.020.503180.9713.49911.415154.13485.1710.062. . .    . . .    2020/2/15 18:47A0019360.010.487175.5113.49911.055149.25985.0430.058          2020/2/17 08:01A0021360.040.499179.613.49911.354153.29585.3540.0592020/2/17 08:32A0022360.020.505181.8113.49911.473154.8885.1880.053. . .    . . .    2020/2/17 18:40A00423600.498179.213.49911.318152.81885.2780.051          . . .    . . .    It is worth noting that although this embodiment only shows the first self-test data corresponding to a first self-test time and the first measurement data corresponding to a first measurement time period, however, in practical applications, It may also be a plurality of first self-test data corresponding to a plurality of first self-test times and a plurality of first measurement data corresponding to a plurality of first measurement time periods, and is not particularly limited. For example, Table 1 below shows examples of a plurality of first self-test data corresponding to a plurality of first self-test times, and Table 2 below shows a plurality of first self-test data corresponding to a plurality of first measurement time periods. Example of first measurement data. Table I Self-test number First self-check time DC voltage V1 DC voltage V2 AC voltage V3 DC voltage V4 DC voltage V5 DC voltage V6 DC current I1 DC voltage V7 DC voltage V8 DC voltage V9 DC current I2 Ground voltage V0 0 2020/2/15 08:01 12.007 12.006 220.010 200.040 200.07 200.04 2.006 350.043 350.05 349.98 1.004 0.003 1 2020/2/17 07:52 12.012 12.009 220.022 200.057 200.08 200.05 2.005 350.086 350.07 350.00 1.004 0.003 2 2020/2/18 07:45 12.011 12.008 220.021 200.067 200.04 200.02 2.005 350.084 350.06 349.98 1.003 0.003 . . . . . .Table II first measurement time Product serial number High voltage voltage (HV Voltage) High voltage current (HV current) Input power (P in) Low voltage voltage (LV Voltage) Low voltage current (LV Current) Output power (P out) Efficient Noise 2020/2/15 08:49 A0001 360.02 0.486 175.11 13.499 10.992 148.405 84.749 0.031 2020/2/15 09:11 A0002 360.02 0.503 180.97 13.499 11.415 154.134 85.171 0.062 . . . . . . 2020/2/15 18:47 A0019 360.01 0.487 175.51 13.499 11.055 149.259 85.043 0.058 2020/2/17 08:01 A0021 360.04 0.499 179.6 13.499 11.354 153.295 85.354 0.059 2020/2/17 08:32 A0022 360.02 0.505 181.81 13.499 11.473 154.88 85.188 0.053 . . . . . . 2020/2/17 18:40 A0042 360 0.498 179.2 13.499 11.318 152.818 85.278 0.051 . . . . . .

如表一所示，量測裝置200可在各自檢時間（即，2020/2/15 08:01、2020/2/17 07:52以及2020/2/18 07:45等時間）進行自我檢測，以產生第一自檢資料（即，包括直流電壓V1、直流電壓V2、交流電壓V3、直流電壓V4、直流電壓V5、直流電壓V6、直流電流I1、直流電壓V7、直流電壓V8、直流電壓V9、直流電流I2以及接地電壓V0等自檢類別的資料）。As shown in Table 1, themeasuring device 200 can perform self-testing at each testing time (i.e., 2020/2/15 08:01, 2020/2/17 07:52, and 2020/2/18 07:45, etc.) , to generate the first self-test data (i.e., including DC voltage V1, DC voltage V2, AC voltage V3, DC voltage V4, DC voltage V5, DC voltage V6, DC current I1, DC voltage V7, DC voltage V8, DC voltage V9, DC current I2 and ground voltage V0 and other self-test category data).

此外，如表二所示，量測裝置200可在多個第一量測時間段（即，表一中的2020/2/15 08:01與2020/2/17 07:52之間的時間段以及2020/2/17 07:52與2020/2/18 07:45之間的時間段等）對機台300(1)~300(N)所製造的多個產品（即，產品序號為A0001~A0042等的產品）進行量測，以產生第一量測資料（即，包括高電壓的電壓（HV Voltage）、高電壓的電流（HV current）、輸入功率（P in）、低電壓的電壓（LV Voltage）、低電壓的電流（LV Current）、輸出功率（P out）、效率（Efficient）以及雜訊（Noise）等量測類別的資料）。In addition, as shown in Table 2, themeasurement device 200 can perform multiple first measurement time periods (ie, the time between 2020/2/15 08:01 and 2020/2/17 07:52 in Table 1 period and the time period between 2020/2/17 07:52 and 2020/2/18 07:45, etc.) for multiple products manufactured by machines 300(1)~300(N) (that is, the product serial number is Products A0001~A0042, etc.) are measured to generate the first measurement data (i.e., including high voltage voltage (HV Voltage), high voltage current (HV current), input power (P in), low voltage Voltage (LV Voltage), low-voltage current (LV Current), output power (P out), efficiency (Efficient), noise (Noise) and other measurement categories).

接著，於步驟S320中，處理器130可產生與第一自檢資料以及第一量測資料相關的預測模型，以儲存於記憶體120中的模型庫122。詳細而言，處理器130可從記憶體120中的資料庫121讀取包括過去所有自檢資料的第一自檢資料以及包括過去所有量測資料的第一量測資料，並利用第一自檢資料以及第一量測資料產生預測模型，以將所產生的預測模型儲存於記憶體120中的模型庫122。Next, in step S320 , theprocessor 130 may generate a prediction model related to the first self-test data and the first measurement data to be stored in themodel library 122 in thememory 120 . Specifically, theprocessor 130 may read the first self-test data including all past self-test data and the first measurement data including all past measurement data from thedatabase 121 in thememory 120, and use the first self-test data. The inspection data and the first measurement data are used to generate a prediction model, and the generated prediction model is stored in themodel library 122 in thememory 120 .

在一些實施例中，處理器130可對第一量測資料以及第一自檢資料執行機器學習演算法以產生預測模型。此機器學習演算法可以是線性迴歸（Linear Regression）演算法、多項式迴歸(Polynomial Regression)演算法或長短期記憶（Long Short-Term Memory，LSTM）演算法等預測分析演算法，並沒有對上述機器學習演算法有特別的限制。In some embodiments, theprocessor 130 may execute a machine learning algorithm on the first measurement data and the first self-test data to generate a prediction model. This machine learning algorithm can be a linear regression (Linear Regression) algorithm, a polynomial regression (Polynomial Regression) algorithm or a long short-term memory (Long Short-Term Memory, LSTM) algorithm and other predictive analysis algorithms. Learning algorithms have special limitations.

在進一步的實施例中，此預測模型可包括多個子模型，其中這些子模型對應於第一量測資料的多個量測類別。詳細而言，處理器130可利用第一自檢資料以及第一量測資料之中與其中一個量測類別對應的資料訓練出與其中一個量測類別對應的子模型。以此類推，處理器130可訓練出與上述這些量測類別對應的多個子模型。In further embodiments, the prediction model may include multiple sub-models, wherein the sub-models correspond to multiple measurement categories of the first measurement data. Specifically, theprocessor 130 may use the data corresponding to one of the measurement categories among the first self-test data and the first measurement data to train a sub-model corresponding to one of the measurement categories. By analogy, theprocessor 130 can train multiple sub-models corresponding to the above-mentioned measurement categories.

舉例而言，如上述表一以及表二，處理器130可計算各第一量測時間段中與HV Voltage（即，上述其中一個量測類別）對應的資料的平均值。換言之，處理器130可計算各第一量測時間段中與HV Voltage對應的資料的平均值（例如，針對HV Voltage，2020/2/15 08:01至2020/2/17 07:52之間所量測的資料的平均值為360.3018，而2020/2/17 07:52至2020/2/18 07:45之間所量測的資料的平均值為360.015）。For example, as shown in Table 1 and Table 2 above, theprocessor 130 can calculate the average value of the data corresponding to the HV Voltage (ie, one of the above measurement categories) in each first measurement time period. In other words, theprocessor 130 can calculate the average value of the data corresponding to the HV Voltage in each first measurement time period (for example, for the HV Voltage, between 2020/2/15 08:01 and 2020/2/17 07:52 The average value of the measured data is 360.3018, and the average value of the measured data between 2020/2/17 07:52 and 2020/2/18 07:45 is 360.015).

藉此，處理器130可對第一自檢資料以及與HV Voltage對應的資料的平均值執行機器學習演算法以產生與HV Voltage對應的子模型。以此類推，處理器130可訓練出分別與這些量測類別對應的多個子模型。Thereby, theprocessor 130 can execute a machine learning algorithm on the average value of the first self-test data and the data corresponding to the HV Voltage to generate a sub-model corresponding to the HV Voltage. By analogy, theprocessor 130 can train multiple sub-models respectively corresponding to these measurement categories.

詳細而言，表三示出了第一自檢資料以及與HV Voltage對應的資料的平均值的對應關係。 表三自檢編號第一自檢時間直流電壓V1直流電壓V2交流電壓V3直流電壓V4直流電壓V5直流電壓V6直流電流I1直流電壓V7直流電壓V8直流電壓V9直流電流I2接地電壓V0與高電壓的電壓（HV Voltage）對應的資料的平均值02020/2/15 08:0112.00712.006220.010200.040200.07200.042.006350.043350.05349.981.0040.003360.301812020/2/17 07:5212.01212.009220.022200.057200.08200.052.005350.086350.07350.001.0040.003360.01522020/2/18 07:4512.01112.008220.021200.067200.04200.022.005350.084350.06349.981.0030.003359.998. . .      . . .       In detail, Table 3 shows the corresponding relationship between the first self-test data and the average value of the data corresponding to the HV Voltage. Table 3 Self-test number First self-check time DC voltage V1 DC voltage V2 AC voltage V3 DC voltage V4 DC voltage V5 DC voltage V6 DC current I1 DC voltage V7 DC voltage V8 DC voltage V9 DC current I2 Ground voltage V0 The average value of the data corresponding to the high voltage voltage (HV Voltage) 0 2020/2/15 08:01 12.007 12.006 220.010 200.040 200.07 200.04 2.006 350.043 350.05 349.98 1.004 0.003 360.3018 1 2020/2/17 07:52 12.012 12.009 220.022 200.057 200.08 200.05 2.005 350.086 350.07 350.00 1.004 0.003 360.015 2 2020/2/18 07:45 12.011 12.008 220.021 200.067 200.04 200.02 2.005 350.084 350.06 349.98 1.003 0.003 359.998 . . . . . .

如表三所示，在2020/2/15 08:01至2020/2/17 07:52之間與HV Voltage對應的資料的平均值可對應於第一自檢時間2020/2/15 08:01，且在2020/2/17 07:52至2020/2/18 07:45之間與HV Voltage對應的資料的平均值可對應於另一第一自檢時間2020/2/17 07:52。以此類推，可計算出與多個第一自檢時間對應的多個平均值。進一步而言，可以這些第一自檢資料作為自變數，並以與HV Voltage對應的資料的這些平均值作為因變數，進而執行機器學習演算法以產生與HV Voltage對應的子模型。以此類推，可訓練出分別與上述多個量測類別對應的多個子模型（即，分別與HV Voltage、HV current、P in、LV Voltage、LV Current、P out、Efficient以及Noise等對應的多個子模型）。As shown in Table 3, the average value of the data corresponding to HV Voltage between 2020/2/15 08:01 and 2020/2/17 07:52 can correspond to the first self-test time 2020/2/15 08: 01, and the average value of the data corresponding to HV Voltage between 2020/2/17 07:52 and 2020/2/18 07:45 can correspond to another first self-test time 2020/2/17 07:52 . By analogy, multiple average values corresponding to multiple first self-test times can be calculated. Furthermore, the first self-test data can be used as independent variables, and the average values of the data corresponding to the HV Voltage can be used as dependent variables, and then a machine learning algorithm can be executed to generate a sub-model corresponding to the HV Voltage. By analogy, multiple sub-models corresponding to the above-mentioned measurement categories can be trained (i.e., multiple sub-models corresponding to HV Voltage, HV current, P in, LV Voltage, LV Current, P out, Efficient and Noise respectively). submodel).

值得注意的是，上述雖以取平均值方法作為例子，然而，在應用面上，更可以取中位數或其他具統計意義的方法對第一量測資料進行處理，以進一步訓練出分別與上述多個量測類別對應的多個子模型。It is worth noting that although the average method is used as an example above, in terms of application, the median or other statistically significant methods can be used to process the first measurement data to further train the respective Multiple sub-models corresponding to the above-mentioned multiple measurement categories.

在進一步的實施例中，機台監控裝置100更可包括顯示器（未繪示）。機台監控裝置100可計算所訓練出的預測模型之模型表現（Model Performance）。如此一來，機台監控裝置100可藉由顯示器顯示所訓練出的預測模型之模型表現，以供使用者對此模型表現進行監控。舉例而言，以由線性迴歸演算法所產生的預測模型為例，機台監控裝置100可計算與此預測模型對應的判定係數（coefficient of determination）的數值，並將此數值顯示於顯示器上。In further embodiments, themachine monitoring device 100 may further include a display (not shown). Themachine monitoring device 100 can calculate the model performance (Model Performance) of the trained prediction model. In this way, themachine monitoring device 100 can display the model performance of the trained prediction model through the display, so that the user can monitor the model performance. For example, taking the prediction model generated by the linear regression algorithm as an example, themachine monitoring device 100 can calculate the value of the coefficient of determination corresponding to the prediction model, and display the value on the display.

接著，於步驟S330中，機台監控裝置100可經由收發電路110接收第二自檢資料，並依據第二自檢資料以利用預測模型產生預測資料。詳細而言，在第一自檢時間之後的第二自檢時間，量測裝置200可進行自我檢測以產生第二自檢資料，並將此第二自檢資料傳送至機台監控裝置100的收發電路110。藉此，機台監控裝置100可從記憶體120中的模型庫122讀取預測模型，並依據第二自檢資料以利用所讀取的預測模型產生預測資料（即，對在第二自檢時間以及第三自檢時間（在第二自檢時間之後）之間（即，第二量測時間）由機台300(1)~300(N)所製造的多個第二產品的量測資料進行預測，以產生此預測資料）。Next, in step S330, themachine monitoring device 100 may receive the second self-test data through thetransceiver circuit 110, and use the prediction model to generate prediction data based on the second self-test data. Specifically, at the second self-test time after the first self-test time, the measuringdevice 200 can perform self-test to generate second self-test data, and transmit the second self-test data to themachine monitoring device 100Transceiver circuit 110. Thereby, themachine monitoring device 100 can read the prediction model from themodel library 122 in thememory 120, and use the read prediction model to generate prediction data based on the second self-test data (ie, for the second self-test) time and the third self-test time (after the second self-test time) (ie, the second measurement time), the measurement of multiple second products manufactured by the machines 300(1)~300(N) data to generate this forecast data).

舉例而言，當機台監控裝置100在第二自檢時間接收第二自檢資料時，機台監控裝置100可進一步依據第二自檢資料以利用預測模型預測在第二量測時間段（第二自檢時間以及第三自檢時間之間的時間段）與上述HV Voltage、HV current、P in、LV Voltage、LV Current、P out、Efficient以及Noise等量測類別對應的資料（類似表二中的其中一列資料）。換言之，機台監控裝置100可預測對在第二量測時間段的第二量測資料進行預測以產生預測資料。For example, when themachine monitoring device 100 receives the second self-test data at the second self-test time, themachine monitoring device 100 can further use the prediction model to predict the second measurement time period ( The time period between the second self-test time and the third self-test time) and the data corresponding to the above measurement categories such as HV Voltage, HV current, P in, LV Voltage, LV Current, P out, Efficient and Noise (similar to the table One of the data in the second column). In other words, themachine monitoring device 100 can predict the second measurement data in the second measurement time period to generate prediction data.

在一些實施例中，機台監控裝置100可依據第二自檢資料以分別利用預測模型所包括的多個子模型進行預測，以產生分別與多個子模型對應的多個子預測資料，進而將這些子預測資料作為預測資料。In some embodiments, themachine monitoring device 100 can perform prediction using multiple sub-models included in the prediction model based on the second self-test data to generate multiple sub-prediction data respectively corresponding to the multiple sub-models, and then combine these sub-models. Forecast data serves as forecast data.

舉例而言，機台監控裝置100可將與第二自檢資料分別輸入與上述HV Voltage、HV current、P in、LV Voltage、LV Current、P out、Efficient以及Noise等量測類別對應的子模型。藉此，與上述HV Voltage、HV current、P in、LV Voltage、LV Current、P out、Efficient以及Noise等量測類別對應的子模型可分別輸出對應的子預測資料。如此一來，機台監控裝置100可將分別從這些子模型輸出的這些子預測資料作為預測資料。For example, themachine monitoring device 100 may input the second self-test data into sub-models corresponding to the above-mentioned measurement categories such as HV Voltage, HV current, P in, LV Voltage, LV Current, P out, Efficient, and Noise. . Thereby, the sub-models corresponding to the above measurement categories such as HV Voltage, HV current, P in, LV Voltage, LV Current, P out, Efficient and Noise can respectively output corresponding sub-prediction data. In this way, themachine monitoring device 100 can use the sub-prediction data output from the sub-models as prediction data.

在一些實施例中，機台監控裝置100可經由該收發電路110接收第二量測資料，其中第二量測相關於由機台300(1)~300(N)所製造的多個第二產品，該第二產品可以是在不同批次製造的相同類型的產品。藉此，當機台監控裝置100判斷與第二量測資料以及預測資料對應的至少一差值（即，第二量測資料中的至少一量測值以及分別與此至少一量測值對應的預測資料中的至少一預測值之間的至少一絕對差值。換言之，當A為預測值，B為量測值時，該差值為|A-B|）未小至少一閾值時，機台監控裝置100可依據第二自檢資料以及第二量測資料更新預測模型。In some embodiments, themachine monitoring device 100 can receive second measurement data via thetransceiver circuit 110 , wherein the second measurement is related to a plurality of second measurements produced by the machines 300(1)˜300(N). product, the second product may be the same type of product manufactured in a different batch. Thereby, when themachine monitoring device 100 determines at least one difference value corresponding to the second measurement data and the prediction data (that is, at least one measurement value in the second measurement data and the at least one measurement value respectively corresponding to At least one absolute difference between at least one predicted value in the prediction data. In other words, when A is the predicted value and B is the measured value, the difference is |A-B|) is not less than at least a threshold, the machine Themonitoring device 100 can update the prediction model based on the second self-test data and the second measurement data.

詳細而言，量測裝置200可對在第二量測時間段（第二自檢時間以及第三自檢時間之間的時間段）由機台300(1)~300(N)所製造的多個第二產品進行量測，以將所產生的第二量測資料傳送至機台監控裝置100。藉此，機台監控裝置100可計算與第二量測資料以及預測資料對應的至少一差值。當機台監控裝置100判斷至少一差值皆未小於至少一閾值時，機台監控裝置100可依據第一自檢資料、第一量測資料、第二自檢資料以及第二量測資料重新訓練出另一預測模型。Specifically, the measuringdevice 200 can measure the products manufactured by the machines 300(1)~300(N) during the second measurement time period (the time period between the second self-test time and the third self-test time). A plurality of second products are measured to transmit the generated second measurement data to themachine monitoring device 100 . Thereby, themachine monitoring device 100 can calculate at least one difference corresponding to the second measurement data and the prediction data. When themachine monitoring device 100 determines that at least one difference value is not less than at least a threshold, themachine monitoring device 100 can restart based on the first self-test data, the first measurement data, the second self-test data and the second measurement data. Train another prediction model.

舉例而言，當機台300(1)~300(N)在第二量測時間段製造10個第二產品時，量測裝置200可對這10個第二產品進行量測以產生與8個量測類別對應的80個子量測資料，以將這80個子量測資料作為第二量測資料。藉此，機台監控裝置100可計算與10個第二產品的各量測類別對應的平均值（例如，若量測類別為HV Voltage，可對由10個第二產品所量測出的與HV Voltage對應的10個資料進行平均值運算，以產生與HV Voltage對應的平均值）。For example, when the machines 300(1)~300(N) manufacture 10 second products during the second measurement time period, themeasurement device 200 can measure the 10 second products to generate 8 The 80 sub-measurement data corresponding to each measurement category are used as the second measurement data. Thereby, themachine monitoring device 100 can calculate the average value corresponding to each measurement category of the 10 second products (for example, if the measurement category is HV Voltage, the average value measured by the 10 second products can be calculated). The 10 data corresponding to HV Voltage are averaged to generate an average value corresponding to HV Voltage).

如此一來，機台監控裝置100可將預測資料中與各量測類別對應的子預測資料以及第二量測資料中與各量測類別對應的平均值進行差值運算，以產生分別與8個量測類別對應的8個差值。因此，機台監控裝置100可判斷這8個差值是否皆小於8個預設的閾值。In this way, themachine monitoring device 100 can perform a difference operation on the sub-prediction data corresponding to each measurement category in the prediction data and the average value corresponding to each measurement category in the second measurement data to generate 8 8 differences corresponding to each measurement category. Therefore, themachine monitoring device 100 can determine whether the eight differences are all less than the eight preset thresholds.

當機台監控裝置100判斷這8個差值皆小於8個預設的閾值時，機台監控裝置100便不會對預測模型進行更新。反之，當機台監控裝置100判斷這8個差值未皆小於8個預設的閾值時，機台監控裝置100可依據第一自檢資料、第一量測資料、第二自檢資料以及第二量測資料重新訓練出另一預測模型。When themachine monitoring device 100 determines that the eight differences are all less than the eight preset thresholds, themachine monitoring device 100 will not update the prediction model. On the contrary, when themachine monitoring device 100 determines that all the eight differences are not less than the eight preset thresholds, themachine monitoring device 100 can based on the first self-test data, the first measurement data, the second self-test data and Another prediction model is retrained with the second measurement data.

值得注意的是，上述雖以取平均值方法作為例子，然而，在應用面上，更可以取中位數或其他具統計意義的方法對第二量測資料進行處理，以進一步判斷是否更新預測模型。It is worth noting that although the average method is used as an example above, in terms of application, the median or other statistically significant methods can be used to process the second measurement data to further determine whether to update the forecast. Model.

在進一步的實施例中，上述至少一閾值可預先儲存於記憶體120中，或者是由使用者預先設定以儲存於記憶體120中。In a further embodiment, the above-mentioned at least one threshold value may be pre-stored in thememory 120 , or may be pre-set by the user to be stored in thememory 120 .

接著，於步驟S340中，機台監控裝置100可判斷預測資料是否在至少一數值範圍之內。當預測資料在數值範圍之內時，進入步驟S350。反之，當預測資料未在數值範圍之內時，進入步驟S360。Next, in step S340, themachine monitoring device 100 may determine whether the predicted data is within at least one numerical range. When the prediction data is within the numerical range, step S350 is entered. On the contrary, when the predicted data is not within the numerical range, step S360 is entered.

詳細而言，至少一數值範圍可分別對應於至少一量測類別。機台監控裝置100可判斷預測資料中與至少一量測類別對應的至少一子預測資料是否皆分別在與至少一量測類別對應的至少一數值範圍之內。換言之，機台監控裝置100可判斷預測資料中與至少一量測類別對應的至少一子預測資料是否皆分別小於上述至少一數值範圍的最大值，且皆分別大於上述至少一數值範圍的最小值。In detail, at least one numerical range may respectively correspond to at least one measurement category. Themachine monitoring device 100 may determine whether at least one sub-prediction data corresponding to at least one measurement category in the prediction data is within at least one numerical range corresponding to at least one measurement category. In other words, themachine monitoring device 100 can determine whether at least one sub-prediction data corresponding to at least one measurement category in the prediction data is each smaller than the maximum value of the above-mentioned at least one numerical range, and each is greater than the minimum value of the above-mentioned at least one numerical range. .

舉例而言，機台監控裝置100可判斷預測資料中與HV Voltage對應的子預測資料(例如，360.5)是否在與HV Voltage對應的數值範圍(例如，359.78至360.86之間)之內。For example, themachine monitoring device 100 may determine whether the sub-prediction data (for example, 360.5) corresponding to the HV Voltage in the prediction data is within a numerical range (for example, between 359.78 and 360.86) corresponding to the HV Voltage.

在一些實施例中，上述數值範圍可以是由機台300(1)~300(N)的廠商提供以儲存於記憶體120中的，也可以是依據第一量測資料以經驗法則(Rule of Thumb)產生的，更可以是由機台管理伺服器(未繪示)產生以儲存於記憶體120中的。In some embodiments, the above numerical range may be provided by the manufacturer of the machines 300(1)~300(N) and stored in thememory 120, or may be based on the first measurement data based on the rule of thumb. Thumb), or may be generated by the machine management server (not shown) and stored in thememory 120.

接著，於步驟S350中，機台監控裝置100可產生與正常狀態(Normal State)相關的監控結果。換言之，機台監控裝置100可產生一個用以指示正常狀態的監控結果。Next, in step S350, themachine monitoring device 100 can generate monitoring results related to the normal state (Normal State). In other words, themachine monitoring device 100 can generate a monitoring result indicating a normal state.

在一些實施例中，當機台監控裝置100產生用以指示正常狀態的監控結果時，機台監控裝置100可在上述顯示器上顯示上述預測結果，以供使用者進行資料監控。In some embodiments, when themachine monitoring device 100 generates a monitoring result indicating a normal state, themachine monitoring device 100 may display the predicted result on the display for the user to perform data monitoring.

最後，於步驟S360中，機台監控裝置100可產生與異常狀態(Abnormal State)相關的監控結果。換言之，機台監控裝置100可產生一個用以指示異常狀態的監控結果。Finally, in step S360, themachine monitoring device 100 can generate monitoring results related to the abnormal state (Abnormal State). In other words, themachine monitoring device 100 can generate a monitoring result indicating an abnormal state.

在一些實施例中，機台監控裝置100可將此監控結果傳送至預警裝置400。藉此，預警裝置400可依據此監控結果判斷是否產生警告訊息。當預警裝置400判斷此監控結果相關於異常狀態，預警裝置400可產生警告訊息，以向使用者警告機台300(1)~300(N)中的任意者或量測裝置200可能已發生異常。In some embodiments, themachine monitoring device 100 can transmit the monitoring results to theearly warning device 400 . Thereby, theearly warning device 400 can determine whether a warning message is generated based on the monitoring results. When theearly warning device 400 determines that the monitoring result is related to an abnormal state, theearly warning device 400 can generate a warning message to warn the user that any one of the machines 300(1)~300(N) or themeasuring device 200 may have an abnormality. .

在進一步的實施例中，預警裝置400可將警告訊息傳送至使用者所使用的任意裝置（例如，智慧型手機、平板電腦、筆記型電腦、桌上型電腦等）以警告使用者，也可以是藉由預警裝置400中的顯示器（未繪示）警告使用者，更可以是將警告訊息傳送至機台監控裝置100，以藉由機台監控裝置100中的顯示器警告使用者 。In further embodiments, thewarning device 400 can send a warning message to any device used by the user (for example, a smart phone, a tablet, a notebook computer, a desktop computer, etc.) to warn the user, or it can The user is warned through the display (not shown) in theearly warning device 400 , or the warning message may be sent to themachine monitoring device 100 to warn the user through the display in themachine monitoring device 100 .

藉由上述步驟，本發明的機台監控裝置100可準確地預測在未來的量測時間段中所製造的產品之量測資料。此外，本發明的機台監控裝置100更可檢測量測裝置200是否發生異常狀態。Through the above steps, themachine monitoring device 100 of the present invention can accurately predict the measurement data of products manufactured in the future measurement time period. In addition, themachine monitoring device 100 of the present invention can further detect whether an abnormal state occurs in themeasuring device 200 .

第4圖是根據本發明一些示範性實施例的機台監控方法的流程圖。同時參照第1圖與第4圖，首先，於步驟S410中，藉由收發電路110接收第一自檢資料以及第一量測資料，其中第一自檢資料相關於量測裝置200，其中第一量測資料相關於由機台300(1)~300(N)所製造的多個第一產品。Figure 4 is a flow chart of a machine monitoring method according to some exemplary embodiments of the present invention. Referring to Figures 1 and 4 at the same time, first, in step S410, the first self-test data and the first measurement data are received through thetransceiver circuit 110, where the first self-test data is related to themeasurement device 200, where the first A measurement data is related to a plurality of first products manufactured by machines 300(1)~300(N).

接著，於步驟S420中，藉由處理器130產生與第一自檢資料以及第一量測資料相關的預測模型。Next, in step S420, theprocessor 130 generates a prediction model related to the first self-test data and the first measurement data.

接著，於步驟S430中，藉由處理器130經由收發電路110接收第二自檢資料，並依據第二自檢資料以利用預測模型產生預測資料。Next, in step S430, theprocessor 130 receives the second self-test data through thetransceiver circuit 110, and uses the prediction model to generate prediction data based on the second self-test data.

最後，於步驟S440中，藉由處理器130依據預測資料以及數值範圍產生監控結果，以依據監控結果進行監控。Finally, in step S440, theprocessor 130 generates monitoring results based on the prediction data and the numerical range, so as to perform monitoring based on the monitoring results.

值得注意的是，本實施例的詳細流程已揭露如上，故不在此進一步贅述。It is worth noting that the detailed process of this embodiment has been disclosed above, so it will not be described further here.

綜上所述，本發明提出的機台監控裝置以及方法可依據量測裝置的過去的自檢資料以及產品的過去的量測資料預測未來的量測資料。此外，更可依據預設的數值範圍以及所預測的量測資料判斷量測裝置是否發生異常狀態。如此一來，可解決無法檢測量測裝置發生異常狀態以造成量測資料失去參考價值的問題。In summary, the machine monitoring device and method proposed by the present invention can predict future measurement data based on past self-test data of the measurement device and past measurement data of the product. In addition, it can be determined whether an abnormal state of the measuring device occurs based on the preset value range and the predicted measurement data. In this way, the problem of being unable to detect an abnormal state of the measurement device and causing the measurement data to lose reference value can be solved.

雖然本發明已以實施例揭露如上，然其並非用以限定本發明，任何所屬技術領域中具有通常知識者，在不脫離本發明的精神和範圍內，當可作些許的更動與潤飾，故本發明的保護範圍當視後附的申請專利範圍所界定者為準。Although the present invention has been disclosed above through embodiments, they are not intended to limit the present invention. Anyone with ordinary knowledge in the technical field may make some modifications and modifications without departing from the spirit and scope of the present invention. Therefore, The protection scope of the present invention shall be determined by the appended patent application scope.

100:機台監控裝置 110:收發電路 120:記憶體 121:資料庫 122:模型庫 130:處理器 200:量測裝置 300(1)~300(N):機台 400:預警裝置 S310~S360、S410~S440:步驟 t、t+1:自檢時間100:Machine monitoring device 110: Transceiver circuit 120:Memory 121:Database 122:Model library 130: Processor 200: Measuring device 300(1)~300(N):Machine 400: Early warning device S310~S360, S410~S440: steps t, t+1: self-test time

第1圖是根據本發明一些示範性實施例的機台監控裝置的方塊圖。 第2圖是根據本發明一些實施例繪示自檢資料以及量測資料的示意圖。 第3圖是根據本發明一些示範性實施例的機台監控方法的示意圖。 第4圖是根據本發明一些示範性實施例的機台監控方法的流程圖。Figure 1 is a block diagram of a machine monitoring device according to some exemplary embodiments of the present invention. Figure 2 is a schematic diagram illustrating self-test data and measurement data according to some embodiments of the present invention. Figure 3 is a schematic diagram of a machine monitoring method according to some exemplary embodiments of the present invention. Figure 4 is a flow chart of a machine monitoring method according to some exemplary embodiments of the present invention.

100:機台監控裝置100:Machine monitoring device

110:收發電路110: Transceiver circuit

120:記憶體120:Memory

130:處理器130: Processor

200:量測裝置200: Measuring device

300(1)~300(N):機台300(1)~300(N):Machine

400:預警裝置400: Early warning device

Claims (10)

Translated fromChinese

一種機台監控裝置，其中該機台監控裝置連接一量測裝置，該量測裝置連接至少一機台，該量測裝置對該至少一機台製造的多個第一產品進行量測，包括：一收發電路，用以接收一第一自檢資料以及一第一量測資料，其中該第一自檢資料為該量測裝置進行自我檢測所產生與該量測裝置相關的資料，其中該第一量測資料為該量測裝置對該至少一機台所製造的多個第一產品進行量測所產生與該多個第一產品相關的資料；一記憶體，用以儲存多個指令；以及一處理器，連接該收發電路以及該記憶體，並用以載入並執行該些指令以：產生與該第一自檢資料以及該第一量測資料相關的一預測模型；經由該收發電路接收一第二自檢資料，並依據該第二自檢資料以利用該預測模型產生一預測資料；以及依據該預測資料以及一數值範圍產生一監控結果，以依據該監控結果進行監控。A machine monitoring device, wherein the machine monitoring device is connected to a measuring device, the measuring device is connected to at least one machine, and the measuring device measures a plurality of first products produced by the at least one machine, including : A transceiver circuit for receiving a first self-test data and a first measurement data, wherein the first self-test data is data related to the measurement device generated by self-test of the measuring device, wherein the first self-test data is The first measurement data is data related to the plurality of first products produced by the measurement device when measuring the plurality of first products manufactured by the at least one machine; a memory for storing a plurality of instructions; and a processor connected to the transceiver circuit and the memory, and used to load and execute the instructions to: generate a prediction model related to the first self-test data and the first measurement data; via the transceiver circuit Receive a second self-test data, and use the prediction model to generate prediction data based on the second self-test data; and generate a monitoring result based on the prediction data and a numerical range, so as to perform monitoring based on the monitoring result.如請求項1所述之機台監控裝置，其中該處理器更用以：對該第一量測資料以及該第一自檢資料執行機器學習演算法以產生該預測模型。The machine monitoring device of claim 1, wherein the processor is further used to: execute a machine learning algorithm on the first measurement data and the first self-test data to generate the prediction model.如請求項1所述之機台監控裝置，其中該第一自檢資料對應於該量測裝置的一第一自檢時間，該第二自檢資料對應於該量測裝置的一第二自檢時間，以及該第一量測資料對應於該量測裝置所量測之該些第一產品的一第一量測時間段，其中該第一量測時間段在該第一自檢時間以及該第二自檢時間之間。The machine monitoring device of claim 1, wherein the first self-test data corresponds to a first self-test time of the measuring device, and the second self-test data corresponds to a second self-test time of the measuring device. inspection time, and the first measurement data corresponds to a first measurement time period of the first products measured by the measurement device, wherein the first measurement time period is during the first self-test time and between the second self-test time.如請求項1所述之機台監控裝置，其中該處理器更用以：經由該收發電路接收一第二量測資料，其中該第二量測資料相關於由該至少一機台所製造的多個第二產品，且該第二量測資料對應於該量測裝置所量測之該些第二產品的一第二量測時間段；以及當判斷與該第二量測資料以及該預測資料對應的至少一差值未小於至少一閾值時，依據該第二自檢資料以及該第二量測資料更新該預測模型。The machine monitoring device of claim 1, wherein the processor is further configured to: receive a second measurement data via the transceiver circuit, wherein the second measurement data is related to a plurality of components produced by the at least one machine. a second product, and the second measurement data corresponds to a second measurement time period of the second products measured by the measurement device; and when judging the second measurement data and the prediction data When the corresponding at least one difference value is not less than at least a threshold value, the prediction model is updated based on the second self-test data and the second measurement data.如請求項1所述之機台監控裝置，其中該處理器更用以：當判斷該預測資料在該數值範圍之內時，產生與一正常狀態相關的該監控結果；以及當判斷該預測資料未在該數值範圍之內時，產生與一異常狀態相關的該監控結果。The machine monitoring device as described in claim 1, wherein the processor is further used to: when it is determined that the predicted data is within the numerical range, generate the monitoring result related to a normal state; and when it is judged that the predicted data is within When the value is not within the range, the monitoring result related to an abnormal state is generated.一種機台監控方法，用於一機台監控裝置中，其中該機台監控裝置連接一量測裝置，該量測裝置連接至少一機台，該量測裝置對該至少一機台製造的多個第一產品進行量測，包括：藉由一收發電路接收一第一自檢資料以及一第一量測資料，其中該第一自檢資料為該量測裝置進行自我檢測所產生與該量測裝置相關的資料，其中該第一量測資料為該量測裝置對該至少一機台所製造的多個第一產品進行量測所產生與該多個第一產品相關的資料；藉由一處理器對該第一量測資料以及該第一自檢資料執行機器學習演算法以產生一預測模型；藉由該處理器經由該收發電路接收一第二自檢資料，並依據該第二自檢資料以利用該預測模型產生一預測資料；以及藉由該處理器依據該預測資料以及一數值範圍產生一監控結果，以依據該監控結果進行監控。A machine monitoring method, used in a machine monitoring device, wherein the machine monitoring device is connected to a measuring device, the measuring device is connected to at least one machine, and the measuring device manufactures multiple products on the at least one machine. Measuring a first product includes: receiving a first self-test data and a first measurement data through a transceiver circuit, wherein the first self-test data is generated by the measurement device during self-test and is related to the quantity. Data related to the measuring device, wherein the first measurement data is data related to the plurality of first products generated by the measuring device measuring the plurality of first products manufactured by the at least one machine; through a The processor executes a machine learning algorithm on the first measurement data and the first self-test data to generate a prediction model; the processor receives a second self-test data through the transceiver circuit, and based on the second self-test data Check the data to generate a prediction data using the prediction model; and use the processor to generate a monitoring result based on the prediction data and a value range to perform monitoring based on the monitoring result.如請求項6所述之機台監控方法，其中該第一自檢資料對應於該量測裝置的一第一自檢時間，該第二自檢資料對應於該量測裝置的一第二自檢時間，以及該第一量測資料對應於該量測裝置所量測之該些第一產品的一第一量測時間段，其中該第一量測時間段在該第一自檢時間以及該第二自檢時間之間。The machine monitoring method of claim 6, wherein the first self-test data corresponds to a first self-test time of the measuring device, and the second self-test data corresponds to a second self-test time of the measuring device. The inspection time, and the first measurement data corresponds to a first measurement time period of the first products measured by the measurement device, wherein the first measurement time period is in the firstbetween the self-test time and the second self-test time.如請求項7所述之機台監控方法，更包括：藉由該處理器經由該收發電路接收一第二量測資料，其中該第二量測資料相關於由該至少一機台所製造的多個第二產品，且該第二量測資料對應於該量測裝置所量測之該些第二產品的一第二量測時間段；以及當藉由該處理器判斷與該第二量測資料以及該預測資料對應的至少一差值未小於至少一閾值時，藉由該處理器依據該第二自檢資料以及該第二量測資料更新該預測模型。The machine monitoring method as claimed in claim 7, further comprising: receiving, by the processor, a second measurement data through the transceiver circuit, wherein the second measurement data is related to a plurality of machines manufactured by the at least one machine. a second product, and the second measurement data corresponds to a second measurement time period of the second products measured by the measurement device; and when the processor determines and the second measurement When at least one difference corresponding to the data and the prediction data is not less than at least a threshold, the processor updates the prediction model based on the second self-test data and the second measurement data.如請求項8所述之機台監控方法，其中藉由該處理器依據該第二自檢資料以及該第二量測資料更新該預測模型的步驟包括：藉由該處理器對該第一量測資料、該第一自檢資料、該第二量測資料以及該第二自檢資料執行機器學習演算法以更新該預測模型。The machine monitoring method of claim 8, wherein the step of updating the prediction model based on the second self-test data and the second measurement data by the processor includes: using the processor to estimate the first quantity The machine learning algorithm is executed on the measurement data, the first self-test data, the second measurement data and the second self-test data to update the prediction model.如請求項6所述之機台監控方法，其中藉由該處理器依據該預測資料以及該數值範圍產生該監控結果，以依據該監控結果進行監控的步驟包括：當藉由該處理器判斷該預測資料在該數值範圍之內時，藉由該處理器產生與一正常狀態相關的該監控結果；以及當藉由該處理器判斷該預測資料未在該數值範圍之內時，藉由該處理器產生與一異常狀態相關的該監控結果。The machine monitoring method as described in claim 6, wherein the processor generates the monitoring result based on the prediction data and the value range, and the step of monitoring based on the monitoring result includes: when the processor determines the When the predicted data is within the numerical range, the processor generates the monitoring result related to a normal state; and when the processor determines that the predicted data is not within the numerical range,The monitoring result related to an abnormal state is generated by the processor.

Images