Automated cleaning machine and method for an automated cleaning machine

Abstract

An automated cleaning machine is provided that includes a trained cleaning outcome classifier that uses machine learning techniques to automatically classify or score cleaning outcomes of the cleaning machine. The cleaning outcome classifier can be trained on training data that includes a plurality of training inputs and a known output for each of the plurality of training inputs. Each of the plurality of training inputs can include one or more cleaning process parameters corresponding to a cleaning process performed by the cleaning machine during a training phase. The known output for each training input can include a cleaning outcome classification or score. A cleaning outcome of a new cleaning process can then be classified or scored with the trained cleaning outcome classifier based on one or more cleaning process parameters corresponding to the new cleaning process.

Description

[0001] This application claims the benefit of U.S. Provisional Application No. 63 / 083,355, filed September 25, 2020, entitled “MACHINE LEARNING CLASSIFICATIONOR SCORING OF CLEANING OUTCOMES IN CLEANING MACHINES”, the entire contents of which are incorporated herein by reference. Background Technology

[0002] Automated cleaning machines are used in restaurants, healthcare facilities, and other locations to clean, sterilize, and / or disinfect a variety of items. In restaurants or food processing facilities, automated cleaning machines (such as dishwashing machines or dishwashers) are used to clean food preparation and dining items, such as tableware, glassware, deep pots, pans, utensils, food processing equipment, and other items. Generally, items to be cleaned are placed on a rack and fed into the cleaning chamber of the automated cleaning machine. In the cleaning chamber, one or more cleaning products and / or rinsing agents are applied to the items during the cleaning process. The cleaning process may include one or more washing stages and one or more rinsing stages. At the end of the cleaning process, the rack is removed from the cleaning chamber. Water temperature, water pressure, water quality, concentration of chemical cleaning and / or rinsing agents, duration of washing and / or rinsing stages, and other factors can affect the effectiveness of the cleaning process. Summary of the Invention

[0003] Generally, this disclosure relates to systems and / or methods for automatically classifying or scoring the cleaning results of cleaning machines using machine learning techniques. For example, a cleaning result classifier can be trained on training data including multiple training inputs and known outputs for each of the multiple training inputs. Each of the multiple training inputs may include one or more cleaning process parameters corresponding to a cleaning process performed by the cleaning machine during the training phase. The known output for each training input may include a cleaning result classification or score. Cleaning process parameters may include one or more of, for example, washing temperature, rinsing temperature, washing time, rinsing time, thermal conductivity of the washing water, detergent type, rinsing aid type, water hardness of the washing water, alkalinity of the washing water, and / or a measurement of the presence of food contaminants in the washing water. The result of the training phase is a trained cleaning result classifier. The cleaning result of a new cleaning process can be classified or scored using this trained cleaning result classifier based on one or more cleaning process parameters corresponding to the new cleaning process.

[0004] In one example, this disclosure relates to an automated cleaning machine comprising: at least one processor; at least one storage device storing one or more predefined cleaning process parameters and a trained cleaning result classifier; the at least one storage device further comprising instructions executable by the at least one processor to: control the cleaning machine to perform at least one cleaning process using the one or more predefined cleaning process parameters; monitor the one or more cleaning process parameters during the execution of the cleaning process; classify or score the result of the cleaning process using the trained cleaning process classifier based on the one or more cleaning process parameters monitored during the execution of the cleaning process; and adjust one or more predefined cleaning process parameters in the predefined cleaning process parameters in response to the trained cleaning process classifier classifying the result of the cleaning process as contaminated, such that subsequent cleaning processes will be classified as clean by the trained cleaning result classifier.

[0005] A trained cleaning process classifier can categorize the outcome of a cleaning process as either clean or contaminated. The trained cleaning process classifier can score the outcome of a cleaning process by assigning numerical scores indicating the cleaning result. One or more cleaning cycle parameters may include one or more of the following: washing temperature, rinsing temperature, washing time, rinsing time, thermal conductivity of the washing water, detergent type, rinsing aid type, water hardness of the washing water, alkalinity of the washing water, and / or a measurement of the presence of food contaminants in the washing water. The measurement of the presence of food contaminants may be a Boolean parameter that sets a first possible value of food contaminants to true and a second possible value of food contaminants to false. The measurement of the presence of food contaminants may also include a measurement of the turbidity of the cleaning solution in the cleaning machine's reservoir.

[0006] The trained cleaning outcome classifier can be either a trained two-level classification machine learning model or a trained regression machine learning model. At least one storage device may also include instructions executable by at least one processor to control the execution of subsequent cleaning processes by the cleaning machine using one or more tuned predefined cleaning process parameters. The trained cleaning outcome classifier can be trained using training data obtained from one or more designed experiments or field tests in which one or more cleaning process validation specimens are placed in the cleaning chamber of the cleaning machine and exposed to cleaning processes performed by the cleaning machine during the training phase. The trained cleaning outcome classifier can be trained based on one or more cleaning process parameters corresponding to each of a plurality of cleaning processes performed during the training phase and known outputs corresponding to each of the plurality of cleaning processes performed during the training phase.

[0007] In another example, this disclosure relates to a method comprising: storing one or more predefined cleaning process parameters and a trained cleaning result classifier in a storage device of an automated cleaning machine; controlling the automated cleaning machine to perform at least one cleaning process using the one or more predefined cleaning process parameters via a controller of the automated cleaning machine; monitoring the one or more cleaning process parameters by the controller of the automated cleaning machine during the execution of the cleaning process; classifying or scoring the result of the cleaning process by the controller of the automated cleaning machine based on the one or more cleaning process parameters monitored during the execution of the cleaning process using the trained cleaning process classifier; and adjusting one or more of the predefined cleaning process parameters by the controller of the automated cleaning machine in response to the trained cleaning process classifier classifying the result of the cleaning process as contaminated, such that subsequent cleaning processes will be classified as clean by the trained cleaning result classifier.

[0008] In another example, this disclosure relates to an automated cleaning machine comprising: at least one processor; at least one storage device storing one or more predefined cleaning process parameters and a trained cleaning result classifier; the at least one storage device further comprising instructions executable by the at least one processor to: control the cleaning machine to perform at least one cleaning process using the one or more predefined cleaning process parameters; monitor one or more cleaning process parameters during the execution of the cleaning process; classify or score the result of the cleaning process using the trained cleaning process classifier based on the one or more cleaning process parameters monitored during the execution of the cleaning process; dynamically adjust one or more predefined cleaning process parameters in response to the trained cleaning process classifier classifying the result of the cleaning process as contaminated, such that the trained cleaning result classifier classifies the cleaning process as clean; and control the cleaning machine to perform the remaining portion of the cleaning process using the dynamically adjusted one or more predefined cleaning process parameters.

[0009] Details of one or more examples are set forth in the accompanying drawings and description below. Other features will be apparent from the description and drawings and from the claims. Attached Figure Description

[0010] Figure 1 An exemplary automated cleaning machine is shown that uses machine learning techniques to automatically classify or score the cleaning results of one or more cleaning processes performed by the cleaning machine, in accordance with this disclosure.

[0011] Figure 2 An exemplary automated cleaning machine according to this disclosure includes one or more cleaning process test pieces for generating training data used to train a cleaning outcome classifier.

[0012] Figures 3A to 3C Exemplary cleaning process test pieces are shown, according to the present disclosure, respectively corresponding to cleaning results that are contaminated, partially contaminated, and clean.

[0013] Figures 3D to 3F Another exemplary cleaning process test piece is shown, according to the present disclosure, corresponding to the classification of cleaning results as contaminated, partially contaminated, and clean, respectively.

[0014] Figure 4 This is a block diagram of an exemplary system according to the present disclosure, wherein an automated cleaning machine uses machine learning techniques to automatically classify or score the cleaning results of one or more cleaning processes performed by the cleaning machine.

[0015] Figure 5 This is a flowchart illustrating an exemplary process according to the present disclosure, through which a computing device can train a cleaning result classifier.

[0016] Figures 6A to 6C This is a graph showing exemplary results obtained from evaluations of different binary cleaning result classifiers and using different feature sets.

[0017] Figure 7 This is a diagram illustrating a summary of exemplary classification model results from several binary classification model tools according to this disclosure.

[0018] Figure 8 This is a graph illustrating a summary of the results of exemplary classification models using several regression modeling tools according to this disclosure.

[0019] Figure 9 This is a flowchart illustrating an exemplary process according to the present disclosure, in which a computing device classifies the results of a cleaning process performed by a cleaning machine using a trained cleaning result classifier.

[0020] Figure 10 This is a flowchart illustrating an exemplary process according to the present invention, in which a computing device uses a trained cleaning process classifier to predict the cleaning result of a new cleaning process and dynamically adjusts one or more cleaning process parameters during the execution of the current cleaning process to ensure a satisfactory cleaning result. Detailed Implementation

[0021] Generally, this disclosure relates to systems and / or methods for automatically classifying or scoring the cleaning results of cleaning machines using machine learning techniques. For example, a cleaning result classifier can be trained on training data including multiple training inputs and known outputs for each of the multiple training inputs. Each of the multiple training inputs may include one or more cleaning process parameters corresponding to a cleaning process performed by the cleaning machine during the training phase. The known output for each training input may include a cleaning result classification or score. Cleaning process parameters may include one or more of, for example, washing temperature, rinsing temperature, washing time, rinsing time, thermal conductivity of the washing water, detergent type, rinsing aid type, water hardness of the washing water, alkalinity of the washing water, and / or a measurement of the presence of food contaminants in the washing water. The result of the training phase is a trained cleaning result classifier. The cleaning result of a new cleaning process can be classified or scored using this trained cleaning result classifier based on one or more cleaning process parameters corresponding to the new cleaning process.

[0022] The cleaning process parameters used to classify or score the results of a new cleaning process can be the same as those used to train the cleaning result classifier during the training phase.

[0023] Training data can be obtained from one or more designed experiments and / or field tests in which one or more cleaning process validation specimens are placed in the cleaning chamber of a cleaning machine and exposed to the cleaning process performed by the cleaning machine. One or more cleaning process parameters are monitored during the execution of the cleaning process, and one or more of these cleaning process parameters are used as training inputs to a cleaning outcome classifier.

[0024] Each validation specimen includes a substrate having at least one test indicator within a validation area of ​​the substrate. When exposed to a cleaning process within a cleaning machine, the test indicator changes, such as being completely removed, partially removed, or changing color. The amount or degree of change is a function of the effectiveness of the cleaning process and is used to assign a known output, such as a cleaning outcome classification or score, to each of a plurality of training inputs. In some examples, to quantify the amount or degree of change in the test indicator as a result of the cleaning process, color and / or grayscale sensor data is obtained from readings of the validation area of ​​the validation specimen. In some examples, a predefined color change threshold can be used to classify a known cleaning output as “clean” or “contaminated.” In other examples, a defined range of color changes can be assigned a range of scores as known outputs.

[0025] A cleaning outcome classifier can be trained on training data, which includes multiple training inputs obtained from designed experiments and / or field tests, and known outputs for each of the multiple training inputs. The cleaning outcome classifier can include any type of machine learning tool, such as a classification tool or a regression tool. A classification-type cleaning outcome classifier can classify each of the multiple training inputs into one of two categories for the known output, such as "clean" or "contaminated." A regression-type cleaning outcome classifier can assign a numerical value or score (e.g., a number from 1 to 100) to the known output. During the training phase, the cleaning outcome classifier uses the training data to find correlations between identified features of the training data that influence the outcome (e.g., one or more cleaning process parameters among the cleaning process parameters). The result of the training phase is the trained cleaning outcome classifier.

[0026] The cleaning results of the new cleaning process can then be classified or scored using the trained cleaning result classifier based on one or more cleaning process parameters corresponding to the new cleaning process. For example, the controller of an automated cleaning machine can be programmed with a trained cleaning result classifier. One or more cleaning process parameters are monitored during the execution of the new cleaning process. One or more of the monitored cleaning process parameters can be used as input to the trained cleaning result classifier to classify or score the cleaning results of the new cleaning process.

[0027] In some examples, the results of a trained cleaning outcome classifier can be used by the cleaning machine controller to automatically adjust one or more cleaning process parameters during subsequent new cleaning processes to ensure a “clean” classification or numerical score associated with a satisfactory cleaning outcome for subsequent new cleaning processes.

[0028] Figure 1 An exemplary automated cleaning machine 100 is shown, which uses machine learning techniques to automatically classify or score the cleaning results of one or more cleaning processes performed by the cleaning machine 100 in accordance with the present disclosure.

[0029] In this example, cleaning machine 100 is a commercial door dishwasher designed for cleaning and / or sterilizing food and / or food preparation articles 102A to 102N. In this example, articles 102A to 102N are plates. However, it should be understood that articles 102A to 102N may also include other food or food preparation articles, such as bowls, coffee cups, glassware, silverware, cooking utensils, pots, pans, etc. It should also be understood that cleaning machine 100 may include any other type of cleaning machine, such as a laundry or textile washing machine, a medical device reprocessor, an automated washing and sterilizing machine, an autoclave, a sterilizer, or any other type of cleaning machine, and this disclosure is not limited to the type of cleaning machine or the type of articles to be cleaned.

[0030] The cleaning machine 100 includes a housing 158 defining one or more cleaning chambers 152 and having one or more doors 160, 161 allowing entry into and / or exit from the cleaning chambers 152. One or more removable racks 154 are sized to fit inside the cleaning chambers 152. Each rack 154 may be configured to receive items to be cleaned directly thereon, or may be configured to receive one or more trays or holders in which the items to be cleaned are held during the cleaning process. The racks 154 may be general-purpose or special-purpose racks and may be configured to accommodate large and / or small items, food processing / preparation equipment such as deep pots, saucepans, cooking utensils, etc., and / or glassware, plates and other eating utensils, etc. In hospital or healthcare applications, the racks may be configured to accommodate instrument trays, durable goods, medical devices, tubing, masks, basins, bowls, bedpans or other medical items. It should be understood that, as per [the relevant regulations], [the following applies]. Figure 1 The configuration of rack 154 and the description of items that may be placed on or in rack 154 shown and described throughout this specification are for illustrative purposes only, and this disclosure is not limited in this respect.

[0031] Typical cleaning machines, such as cleaning machine 100, operate by spraying one or more cleaning solutions 164 (a mixture of water and one or more chemical cleaning products) into a cleaning chamber 152 and thus onto the workpiece to be cleaned. The cleaning solution is pumped to one or more spray arms 162, which spray the cleaning solution 164 into the cleaning chamber 152 at appropriate times. Cleaning machine 100 is provided with a fresh water source and, depending on the application, may also include one or more storage tanks, such as storage tank 110, to hold used cleaning and / or rinsing solutions 112 that will be reused in the next cleaning cycle. Cleaning machine 100 may also include or be provided with a chemical product dispenser 240 that automatically dispenses appropriate chemical products at appropriate times during the cleaning process, mixes them with a diluent, and distributes the resulting cleaning solution 164 into cleaning machine 100 for distribution into the cleaning chamber 152. Depending on the machine, the products to be cleaned, the amount of dirt on the products to be cleaned, and other factors, one or more washing stages may be alternated with one or more rinsing stages and / or disinfection stages to form a complete cleaning process for cleaning machine 100.

[0032] The automated cleaning machine 100 further includes a cleaning machine controller 200. The controller 200 includes one or more processors and / or processing circuits for monitoring and controlling various cleaning process parameters of the cleaning machine 100, such as washing temperature, tank temperature, rinsing temperature, washing and rinsing time and sequence, cleaning solution concentration, timing of dispensing one or more chemical products, amount of chemical products to be dispensed, and timing of applying water and chemical products to the cleaning chamber. The controller 200 can communicate with a product dispensing system 240 to monitor and / or control the timing and / or amount of cleaning products dispensed into the cleaning machine 100.

[0033] In some examples, the cleaning machine controller 200 and / or product distribution system 240 may be configured to communicate with one or more remote computing devices or cloud-based server computing systems (see, for example...). Figure 4 The cleaning machine controller 200 and / or product dispensing system 240 can also be configured to communicate directly or remotely with one or more user computing devices, such as tablet computers, mobile computing devices, smartphones, laptop computers, etc.

[0034] like Figure 1As shown, one or more items to be cleaned, such as plates 102A to 102N, can be placed on rack 154 and moved into cleaning chamber 152 at the start of the cleaning process. Rack 154 can move on conveyor belt 166 or other support structure. Cleaning machine 100 may include one or more sensors that monitor one or more cleaning process parameters during each cleaning process. For example, cleaning machine 100 may include one or more temperature sensors 153 that measure the temperature inside cleaning chamber 152. Figure 1 In the example, temperature sensor 153 is positioned on a side wall inside the cleaning chamber 152 of the cleaning machine 100. The cleaning machine 100 may also include an inlet water supply temperature sensor 151 that measures the temperature of fresh rinse water delivered to the cleaning chamber of the cleaning machine 100. The cleaning machine 100 may also include a reservoir temperature sensor 114 that measures the temperature of solution 112 in reservoir 110. For example, the reservoir water temperature may be measured at the beginning of a cleaning cycle and at the end of the same cleaning cycle to determine differences in reservoir water temperature that occur during the cleaning cycle. As another example, the reservoir water temperature may be measured or sampled at periodic intervals or continuously at predetermined times during the cleaning cycle throughout the cleaning cycle. As another example, temperature sensors (such as temperature sensor 155) may be located at one or more locations on the rack or at one or more locations on a base that holds the rack in the cleaning chamber to measure the water temperature at those locations.

[0035] According to this disclosure, the controller 200 of the cleaning machine 100 uses machine learning techniques according to this disclosure to automatically classify or score the cleaning results of one or more new cleaning processes performed by the cleaning machine 100. For example, the controller 200 of the cleaning machine 100 may monitor one or more cleaning process parameters during the execution of a new cleaning process, and may use a trained cleaning process classifier to classify or score the cleaning results of the new cleaning process.

[0036] The controller 200 can use the cleaning results of a new cleaning process to adjust one or more cleaning process parameters during subsequent new cleaning processes to ensure a "clean" cleaning result or a score associated with a satisfactory cleaning result for the subsequent new cleaning process. For example, when the cleaning result of a new cleaning process is classified as "contaminated" or assigned a score associated with an unsatisfactory cleaning result, the controller 200 can adjust one or more cleaning process parameters until a trained cleaning process classifier predicts a "clean" result or another cleaning score associated with a satisfactory cleaning result, and the adjusted cleaning process parameters can be used to ensure a satisfactory cleaning result during subsequent new cleaning processes.

[0037] In another example, controller 200 may use a trained cleaning process classifier to dynamically adjust one or more cleaning process parameters of the current new cleaning process to ensure a satisfactory cleaning result. Controller 200 may monitor one or more cleaning process parameters during the execution of the current new cleaning process. Controller 200 may, at one or more times during the current new cleaning process and using the trained cleaning process classifier, predict the cleaning result of the current new cleaning process based on one or more monitored cleaning process parameters. Controller 200 may use this prediction to dynamically adjust one or more cleaning process parameters of the current new cleaning process to ensure a "clean" cleaning result or a score associated with a satisfactory cleaning result for the current new cleaning process. For example, if the cleaning result of the current new cleaning process is predicted as "contaminated" or assigned a score associated with an unsatisfactory cleaning result, controller 200 may dynamically adjust one or more cleaning process parameters during the execution of the current cleaning process so that the trained cleaning process classifier predicts a "clean" result or other cleaning score associated with a satisfactory cleaning result for the current new cleaning process. In this way, the number of cleaning processes with unsatisfactory cleaning results can be reduced because the cleaning process parameters associated with the current cleaning process can be dynamically adjusted during the execution of the current cleaning process itself to ensure satisfactory cleaning results are achieved.

[0038] In some examples, the cleaning machine controller 200 or remote computing system (see, for example) Figure 4 The controller 200 can generate one or more reports or notifications about cleaning results determined by a trained cleaning result classifier. For example, the controller 200 can generate a notification for display, such as on a user's computing device, based on the cleaning results generated by the trained cleaning result classifier. This notification includes a cleaning result classification or score assigned to the cleaning process by the trained cleaning result classifier. The displayed data may also include one or more graphs or charts about the data monitored or generated during the cleaning process.

[0039] Figure 2 The following is shown in accordance with this disclosure: Figure 1An exemplary automated cleaning machine 100 of the type shown includes one or more cleaning process test pieces 180A to 180C (generally referred to as validation test pieces 180) for generating training data to train a cleaning outcome classifier. The training data may be obtained during the training phase from one or more designed experiments and / or field tests in which one or more cleaning process validation test pieces 180A to 180C are placed in the cleaning chamber 152 of the cleaning machine 100 and exposed to the cleaning process performed by the cleaning machine 100. One or more cleaning process parameters are monitored during the execution of the cleaning process, and one or more of these cleaning process parameters are used as training inputs to the cleaning outcome classifier. While in Figure 2 Three validation test pieces 180A to 180C are shown, but it should be understood that one or more clean validation test pieces 180 may be used, and the validation test pieces 180 may be placed in different locations within or above the rack 154, and this disclosure is not limited in this respect.

[0040] Figures 3A to 3C It shows the time before exposure to the cleaning process ( Figure 3A Partially contaminated after exposure to the cleaning process. Figure 3B ), and being cleaned after exposure to the cleaning process ( Figure 3C An exemplary cleaning process test piece 180 is provided. The test piece 180 includes a substrate 186 with a test indicator 184 within a verification area 182. When exposed to a cleaning process within a cleaning machine, the test indicator 184 changes, such as being completely removed, partially removed, or changing color. For example, Figure 3B It shows Figure 3A An exemplary cleaning process verification test piece 180, wherein the test indicator 182 has been partially removed during the cleaning process, and Figure 3C It shows Figure 3A An exemplary cleaning process verification test piece 180 is provided, wherein the test indicator 182 has been completely removed by the cleaning process.

[0041] Test indicators may include a single indicative stain, such as Figures 3A to 3C As shown, or may include multiple indicative contaminants. For example, the test indicator may include more than one type of contaminant within the verification area 182, and / or may include the level of more than one type of contaminant within the verification area 182. The type of cleaning process test piece 180 to be used and / or the type of test indicator 184 may depend on one or more of the following, such as the specific application or customer, the type of cleaning machine, the vessel to be cleaned, the type of contaminant that may be encountered in the application.

[0042] The amount or degree of change is a function of the cleaning process's effectiveness and is used to assign known outputs, such as cleaning outcome classifications or scores, to each cleaning process performed during the training phase. To quantify the amount or degree of change as a test indicator of the cleaning process outcome, color and / or grayscale sensor data can be obtained from readings of the verification area of ​​a verification test piece. In some examples, predefined thresholds can be used to classify known cleaning outputs as "clean" or "contaminated." In other examples, a defined range of color changes can be assigned a range of scores as known cleaning outputs.

[0043] The substrate 186 may comprise any type of temperature-stable material, such as plastic, paper, metal, or ceramic. Examples of suitable substrate materials include, but are not limited to, polyethylene, polypropylene, polyester, polyvinyl chloride (vinyl), high-density polyethylene (HDPE), polyethylene terephthalate (PET), and synthetic forms of paper, plastic, ceramic, stainless steel, and other metals. The test indicator 184 may be printed, inkjet printed, screen printed, sprayed, dip-coated, or otherwise deposited on the substrate 186.

[0044] The validation test piece 180 may also include one or more other areas, such as a writable area 188, which allows the user to add identification information or other annotations to the validation test piece 180. Identification information may include, for example, the date and time of the cleaning cycle, the identification of the cleaning machine, the identification of the personnel running the cleaning cycle and / or validation procedure, a "clean" or "contaminated" indication, and / or other information related to the cleaning process validation procedure. The validation test piece 180 may also include a printed identifier 190 that uniquely identifies the test piece. Figures 3A to 3C In the example, identifier 190 is a human visually readable and / or computer device electronically readable serial number. In other examples, identifier 190 may also include one or more of barcodes, QR codes, or other types of electronically readable identifiers or codes.

[0045] Each validation test piece 180 and test indicator 184 is designed to represent dirt experienced in a specific application and in response to a cleaning process suitable for those applications. For example, in restaurants or other food establishments, automated cleaning machines may include automated dishwashers, and the cleaning process is expected to remove food and / or other dirt typically encountered in such applications. Therefore, test indicators designed for such applications may include food-based dirt such as fats and oils, proteins, carbohydrates, food dyes, minerals, starches, coffee and tea stains, etc., or other dirt typically encountered in food establishments such as dyes, inks, lipsticks, or other cosmetic stains. In healthcare applications, test indicators may include dirt typically found in or representing dirt encountered in a medical setting, which may also include organic dirt such as proteins, lipids, carbohydrates, bone fragments, etc., and / or inorganic dirt such as saline solution, dimethyl silicone oil, bone cement, calcium and other minerals, dyes, inks, etc. In other applications, test indicators may include dirt or stains typically found in or representing dirt encountered in such applications, and this disclosure is not limited in this respect.

[0046] Figures 3D to 3F Another exemplary cleaning process test piece 192 is shown, corresponding to cleaning results categorized according to this disclosure as contaminated, partially contaminated, and clean. In this example, the test piece 192 includes a substrate 193 with three test indicators 196A to 196C within a verification area 194. The test indicators 196A to 196C consist of three distinct types of engineered dirt with varying degrees of removal difficulty. Differences between test indicators 196A to 196C may include, for example, the color of the engineered dirt, the size and / or geometry of the dirt spots, and / or the composition of the engineered dirt. The test piece 192 thus presents three distinct challenges to the cleaning process. When exposed to the cleaning process within the cleaning machine, the test indicators 196A to 196C change, such as complete removal, partial removal, or color change. The type of the cleaning process test piece 192 and / or the quantity and / or type of the test indicators 196A to 196C to be used may depend on one or more of, for example, the specific application or customer, the type of cleaning machine, the vessel to be cleaned, the type of dirt that may be encountered in the application, etc. Although the validation test piece 192 is shown and described as comprising three distinct dirtes, it should be understood that the validation test piece may comprise a single dirt, two distinct dirtes, or three or more distinct dirtes, and this disclosure is not limited in this respect. It should also be understood that any other variation of the exemplary validation test piece 192 or the validation test piece may be as follows: Figure 2 It may be used in place of or in combination with the exemplary verification test piece 180 in the cleaning machine shown or otherwise described herein.

[0047] Figures 3D to 3FIt shows the time before exposure to the cleaning process ( Figure 3D Partially contaminated after exposure to the cleaning process. Figure 3E ), and being cleaned after exposure to the cleaning process ( Figure 3F Exemplary cleaning process test piece 192. For example, Figure 3E It shows Figure 3D An exemplary cleaning process validation test piece 192 is provided, wherein test indicators 196A, 196B and 196C have been partially removed by the cleaning process, but to varying degrees due to their different types of dirt and / or different levels of removal difficulty. Figure 3F It shows Figure 3A An exemplary cleaning process is used to verify test piece 192, wherein each of test indicators 196A to 196C has been completely removed by the cleaning process.

[0048] The amount or degree of change is a function of the cleaning process's effectiveness and is used to assign known outputs, such as cleaning outcome classifications or scores, to each cleaning process performed during the training phase. To quantify the amount or degree of change as a test indicator of the cleaning process outcome, color and / or grayscale sensor data can be obtained from readings of the verification area of ​​a verification test piece. In some examples, predefined thresholds can be used to classify known cleaning outputs as "clean" or "contaminated." In other examples, a defined range of color changes can be assigned a range of scores as known cleaning outputs.

[0049] Substrate 193 may comprise any type of temperature-stable material, such as plastic, paper, metal, or ceramic. Examples of suitable substrate materials include, but are not limited to, polyethylene, polypropylene, polyester, polyvinyl chloride (vinyl), high-density polyethylene (HDPE), polyethylene terephthalate (PET), and synthetic forms of paper, plastic, ceramic, stainless steel, and other metals. Test indicators 196A to 196C may be printed, inkjet printed, screen printed, sprayed, dip-coated, or otherwise deposited on substrate 193. Test indicators 196A to 196C may be deposited on substrate 193 using the same or different manufacturing techniques.

[0050] The validation test piece 192 may also include one or more other areas, such as writable areas, which allow the user to add identification information or other annotations to the validation test piece 192. The writable areas may be on the front of the validation test piece 192 or on the back of the validation test piece 192 (not shown). Identification information may include, for example, the date and time of the cleaning cycle, the identification of the cleaning machine, the identification of the personnel running the cleaning cycle and / or validation procedure, a "clean" or "contaminated" indication, and / or other information related to the cleaning process validation procedure. The validation test piece 192 may also include a printed identifier that uniquely identifies the test piece. For example, similar to... Figures 3A to 3C The identifier 190 and test piece 192 may also include identifiers such as serial numbers that are visually readable by humans and / or electronically readable by computing devices. In other examples, similar to... Figures 3A to 3C The described example may also include one or more of barcodes, QR codes, or other types of electronically readable identifiers or codes.

[0051] Each validation test piece 192 and test indicators 196A to 196C are designed to represent dirt experienced in a specific application and in response to a cleaning process suitable for those applications. For example, in restaurants or other food establishments, automated cleaning machines may include automated dishwashers, and the cleaning process is expected to remove food and / or other dirt commonly encountered in such applications. Therefore, test indicators designed for such applications may include food-based dirt such as fats and oils, proteins, carbohydrates, food dyes, minerals, starches, coffee and tea stains, etc., or other dirt commonly encountered in food establishments such as dyes, inks, lipsticks, or other cosmetic stains. In healthcare applications, test indicators may include dirt commonly found in or representing dirt encountered in a medical setting, which may also include organic dirt such as proteins, lipids, carbohydrates, bone fragments, etc., and / or inorganic dirt such as saline, dimethyl silicone oil, bone cement, calcium and other minerals, dyes, inks, etc. In other applications, test indicators may include dirt or stains commonly found in or representing dirt encountered in such applications, and this disclosure is not limited in this respect. For validation test piece 192, three unique test indicators 196A to 196C can be included to suit any three different types of dirt challenges for the application.

[0052] See you again Figure 2 One or more cleaning process parameters are monitored during the execution of each cleaning process during the training phase. Once the cleaning machine has completed the execution of the cleaning process during the training phase, validation pieces, such as validation pieces 180, 192, or other types of validation pieces associated with the cleaning process, are removed from cleaning machine 150. The one or more cleaning process parameters monitored during the execution of the cleaning process form the training inputs to the cleaning outcome classifier. The amount of dirt retained on the validation pieces indicates the effectiveness of the cleaning process. The amount of dirt retained on the validation pieces can be quantified to assign a known output to each training input.

[0053] In one example, to quantify the amount of dirt remaining on verification test pieces 180, 192, or other exemplary verification test pieces, a color sensor can be used to obtain a color reading associated with a verification area (e.g., verification area 182 of test piece 180 or verification area 194 of verification test piece 192) after the cleaning process is complete. The color reading can be transmitted to a computing device (see, for example...). Figure 4 The computing device receives and analyzes color readings to generate additional color data. The color data may include, for example, one or more RGB ratios. RGB ratios may include, for example, red / green ratios (R / G), red / blue ratios (R / B), and / or blue / green (B / G) ratios. Additionally or alternatively, the color data may include one or more percentage color values. Percentage color values ​​may include, for example, percentage red (%R), percentage blue (%B), and / or percentage green (%G). The color data may also include FIJI grayscale values. Other color data may also be generated, and this disclosure is not limited in this respect. For a validation test piece, such as test piece 192 having one or more test indicators (such as test indicators 196A to 196C within validation area 194), the color data may include separate color data associated with each of the test indicators 196A to 196C within validation area 194.

[0054] In some examples, if certain contaminants remain, such as proteins (coomassie blue or silver staining), carbohydrates, fats, blood, etc., the test indicator can be stained or dyed to cause a color change. Staining or dyeing the test indicator can help make certain changes in the test indicator easier to detect in some cases.

[0055] Exemplary techniques for quantifying the amount of dirt remaining on a verification test piece after a cleaning process has been completed are described in U.S. Provisional Application No. 62 / 942,801, filed December 3, 2019, entitled “Verification of Cleaning Process Efficacy,” the entire contents of which are incorporated herein by reference. However, it should be understood that other techniques for quantifying the amount of dirt remaining on a verification test piece may also be used, and this disclosure is not limited in this respect. In an alternative example, the amount of remaining dirt, or whether a verification test piece should be classified as “clean” or “contaminated,” may be determined manually by visual inspection.

[0056] It should also be understood that other cleaning process validation techniques for determining the effectiveness of a cleaning process can replace the cleaning process validation strips described herein, and such alternative cleaning process validation techniques can be used to train machine learning models for classifying or scoring cleaning results as described herein, and this disclosure is not limited to this aspect. For example, grading systems based on visual inspection of glassware or validation strips, measurement of residual bacterial growth, protein staining, ATP wiping, and bioluminescence measurements to detect residual ATP as an indicator of surface cleanliness can also be used to determine and / or measure the effectiveness of a cleaning process.

[0057] To assign known cleaning outputs to each cleaning process, in some examples, a predefined color change threshold can be used to classify the known cleaning output as "clean" or "contaminated." In other examples, a defined range of color changes can be assigned a range of scores as known cleaning outputs. For example, Figure 3A and Figure 3B The exemplary verification test piece in the sample will be classified as "contaminated," while Figure 3C The exemplary sample will be classified as "clean". As another example, Figure 3D and Figure 3E The exemplary verification test piece 192 can be classified as "contaminated" (i.e., a Boolean value of clean or contaminated) or classified as different levels of "contamination" (e.g., a score from 1 to 5, where 1 is the least contaminated and 5 is the most contaminated, or some other user-defined scoring method), while Figure 3F Exemplary test piece 192 can be classified as "clean".

[0058] A cleaning outcome classifier can be trained on training data, which includes multiple training inputs obtained from designed experiments and / or field tests, and known outputs for each of the multiple training inputs. The cleaning outcome classifier can include any type of machine learning tool, such as a classification tool or a regression tool. A classification tool can classify each of the multiple training inputs into one of several categories of known outputs, such as "clean" or "contaminated." A regression tool can quantize each of the multiple training inputs into a value or score (e.g., a number from 1 to 100) of the known output. The cleaning outcome classifier uses the training data to find correlations between identified features of the training data that influence the outcome (e.g., one or more cleaning process parameters among the cleaning process parameters).

[0059] The cleaning results of the new cleaning process can then be classified or scored using the trained cleaning result classifier based on one or more cleaning process parameters corresponding to the new cleaning process. For example, automated cleaning machines (such as...) Figure 1 The controller of the cleaning machine 100 can be programmed with a trained cleaning result classifier. One or more cleaning process parameters are monitored during the execution of a new cleaning process. These monitored parameters are used as input to the trained cleaning result classifier to classify or score the new cleaning process.

[0060] Additionally, in some examples, the results of a trained cleaning outcome classifier can be used by the cleaning machine controller to automatically adjust one or more cleaning process parameters during subsequent new cleaning processes to ensure a “clean” classification or numerical score associated with satisfactory cleaning results in subsequent new cleaning processes.

[0061] Figure 4 This is a block diagram illustrating an exemplary cleaning machine controller 200 according to the present disclosure, which uses machine learning techniques to analyze the cleaning machines (such as...) of an associated cleaning machine. Figure 1 The cleaning results of one or more new cleaning processes performed by the cleaning machine 100 shown are automatically classified or scored. For example, the controller 200 includes a trained cleaning result classifier 218 that classifies or scores the cleaning results of one or more new cleaning processes performed by the associated cleaning machine.

[0062] The cleaning machine controller 200 is a computing device that includes one or more processors 202, one or more user interface components 204, one or more communication components 206, and one or more data storage components 210. The user interface component 204 may include one or more audio interfaces, visual interfaces, and touch-based interface components, including touch-sensitive screens, displays, speakers, buttons, keyboards, styluses, mice, or other mechanisms that allow human interaction with the computing device. The communication component 206 allows the controller 200 to communicate with other electronic devices, such as a product dispenser controller 242 and / or other remote or local computing devices 250. This communication can be accomplished via wired and / or wireless communication, as typically indicated by network 230.

[0063] Controller 200 includes one or more storage devices 208, which include a cleaning process control module 212, stored cleaning cycle parameters 214, a trained cleaning result classifier 218, an analysis / reporting module 216, and a data storage device 210. Modules 212, 216, and 218 can perform the described operations using software, hardware, firmware, or a mixture of hardware, software, and firmware residing in and / or executing at controller 200. Controller 200 can execute modules 212, 216, and / or 218 using one or more processors 202. Controller 200 can execute modules 212, 216, and / or 218 as a virtual machine executing on the underlying hardware. Modules 212, 216, and / or 218 can be executed, for example, as a service or component of an operating system or computing platform by one or more remote computing devices 250. Modules 212, 216, and / or 218 can be executed as one or more executable programs at the application layer of the computing platform. User interface 204 and modules 212, 216 and / or 218 may be additionally remotely arranged in and accessible by controller 200, for example, as one or more network services operating in a cloud-based network computing system provided by one or more remote computing devices in remote computing device 250.

[0064] Cleaning cycle parameters 214 include cleaning process parameters for one or more default cleaning cycles, such as "Normal," "Deep Pot / Flat Pan," "Heavy Load," etc. Cleaning process parameters may include, for example, the timing and sequence of washing and rinsing phases, washing and rinsing water temperature, tank water temperature, washing and rinsing water conductivity, washing phase duration, rinsing phase duration, residence time duration, washing and rinsing water pH, detergent concentration, rinsing agent concentration, humidity, water hardness, turbidity, rack temperature, mechanical motion within the cleaning machine, and any other cleaning process parameters that may affect the effectiveness of the cleaning process. For each type of cleaning cycle, the values ​​of one or more cleaning process parameters may differ. For example, compared to a "Normal" cleaning cycle, a "Heavy Load" cleaning cycle may include one or more of the following cleaning process parameters: higher washing water temperature, higher rinsing water temperature, longer washing time, larger amount of cleaning product, or other different cleaning cycle parameters. Depending on the type of machine, the cleaning process parameters may differ; for example, gantry machines and conveyor machines may have different cleaning process parameters.

[0065] The cleaning process control module 212 includes instructions executable by the processor 202 to perform various tasks. For example, the cleaning process control module 212 includes instructions executable by the processor 202 to initiate and / or control one or more new cleaning processes in associated cleaning machines. The controller 200 also monitors one or more cleaning process parameters during the execution of the cleaning process. Cyclic data corresponding to each cleaning process performed by the cleaning machines (including one or more cleaning process parameters monitored during the execution of the cleaning process or otherwise corresponding to the cleaning process) can be stored in the data storage device 210.

[0066] According to this disclosure, the trained cleaning result classifier 218 includes instructions executable by processor 202 to classify the cleaning results of associated cleaning machines (such as...) using machine learning techniques according to this disclosure. Figure 1 The cleaning results of the cleaning process performed by the cleaning machine 100 shown are automatically classified or scored. For example, the cleaning results of a new cleaning process performed by the cleaning machine can then be classified or scored using a trained cleaning result classifier 218 based on one or more cleaning process parameters monitored during the new cleaning process or otherwise associated with the new cleaning process.

[0067] The analysis / reporting module 216 (or any of the cleaning process control modules 212, or other software or modules stored in the storage device 208) can generate one or more notifications or reports about the cleaning results of one or more new cleaning cycles for storage or display on the user interface 204 of the controller 200, or on any other local or remote computing device 250.

[0068] As another example, reports may include data associated with cleaning processes performed at a specific cleaning machine, a group of one or more cleaning machines, a cleaning machine at a specific location or group of locations, or a cleaning machine associated with a specific company entity or group of entities. Reports may also include data associated with cleaning processes performed by date / time, by employee, etc. This data can be used to identify trends, areas for improvement, or otherwise help organizational personnel responsible for ensuring the effectiveness of the cleaning cycle to identify and address problems with the cleaning machines.

[0069] The report may also include information monitored during one or more cleaning processes, and the data for each cleaning process may include information monitored during the execution of that cleaning process, such as the date and time of the cleaning process, the unique identifier of the cleaning machine, the unique identifier of the person operating the cleaning process, the type of products cleaned during the cleaning process, the rack capacity or type of rack or tray used during the cleaning process, the duration of the washing phase, the duration of the rinsing phase, the residence time, the temperature of the washing and rinsing water, the temperature of the tank water, the conductivity of the washing and rinsing water, the pH of the washing and rinsing water, the detergent concentration, the rinsing agent concentration, the ambient humidity, the water hardness, the turbidity, the presence / absence of food contaminants in the tank, the rack temperature, the type and amount of chemical products dispensed during each stage of the cleaning process, the volume of water dispensed during each stage of the cleaning process, the total amount of heat equivalents (HUE) accumulated during the cleaning cycle, or other information related to the cleaning process. The report may also include information about location; business entity / enterprise; company cleaning validation targets and tolerances; cleaning score based on location, area, machine type, date / time, employees and / or type of cleaning chemicals; energy costs; chemical product costs; water consumption; and / or any other cleaning cycle data collected or generated by the system or requested by the user.

[0070] Figure 5 This is a flowchart illustrating an exemplary process (300) for a training phase according to the present disclosure, wherein a computing device trains a clean results classifier during the training phase. The computing device may include, for example... Figure 4 Any of the exemplary computing devices 250, and the process (300) may be controlled at least in part based on the execution of instructions stored in the machine learning tool and executed by the processor 252.

[0071] In this example, a cleaning outcome classifier is trained on training data, which includes multiple training inputs and a known output for each of the multiple training inputs. Each of the multiple training inputs corresponds to a cleaning process performed by a cleaning machine during the training phase. The cleaning process performed during the training phase may be performed by one or more cleaning machines. The known output for each training input may include a cleaning outcome classification or score. Each training input corresponding to a cleaning process performed during the training phase may include one or more cleaning process parameters, such as those monitored during the execution of the cleaning process or otherwise corresponding to the cleaning process. Cleaning process parameters may include one or more of the following: washing temperature, rinsing temperature, washing time, rinsing time, thermal conductivity of the washing water, detergent type, rinsing aid type, water hardness of the washing water, alkalinity of the washing water, and / or a measurement of the presence of food dirt in the washing water. The result of the training phase is a trained cleaning outcome classifier that classifies or scores the cleaning outcome of the new cleaning process based on one or more cleaning process parameters monitored during the new cleaning process or otherwise corresponding to the new cleaning process.

[0072] At the start of process (300), the computing device receives training data (302). The training data includes multiple training inputs, each of which has a corresponding known training output. Each of the multiple training inputs and the corresponding known training output is associated with a different cleaning process among multiple cleaning processes performed by one or more cleaning machines during the training phase. The training data may be obtained from one or more designed experiments and / or field tests in which one or more cleaning process validation specimens (or other means for validating the effectiveness of the cleaning process) are placed in the cleaning chamber of the cleaning machine and exposed to the cleaning processes performed by the cleaning machine. During the training phase, one or more cleaning process parameters are monitored during the execution of the cleaning process, and a subset of one or more of these cleaning process parameters is used as training input to a cleaning outcome classifier.

[0073] The known training output can be, for example, a binary classification (e.g., "clean" or "contaminated"). In other examples, the known training output can be a quantified or numerical score (e.g., a score from 0 to 100 or some other numerical range) that indicates the relative amount of dirt retained on the validation test piece and thus indicates the relative effectiveness of the cleaning process.

[0074] At step (304), a “feature set” of one or more cleaning process parameters to be used to train the cleaning outcome classifier is selected. For example, based on analysis performed on the training data using one or more machine learning tools according to this disclosure, one or more cleaning process parameters may be identified as relatively more important for predicting cleaning outcomes than other parameters. Additionally, certain combinations of one or more cleaning process parameters may be identified as relatively more important for predicting cleaning outcomes. The selection of the feature set may also be based on which cleaning process parameters are measured or available. Examples of different feature sets for cleaning process parameters are shown in the following table.

[0075] Table 1:

[0076] Table 1

[0077]

[0078] Once the cleaning process parameters (304) used as input to the cleaning result classifier are selected, the selected training data is divided into a first training data subset to be used to train the cleaning result classifier and a second training data subset to be used to evaluate the trained cleaning result classifier generated based on the first training data subset (306).

[0079] The cleaning outcome classifier can be implemented using any type of machine learning algorithm or tool, such as a binary classification model or a regression model (308). Examples of different machine learning tools include logistic regression (LR), linear regression, boosting decision trees, Bayesian point machines, Naive Bayes, random forests (RF), neural networks (NN), and support vector machines (SVM). In some examples, the tool can be implemented as a two-level binary classification model (e.g., "clean" or "contaminated") or a regression model that generates a quantified numerical score indicating the relative amount of dirt remaining on the validation test piece and thus indicating the relative effectiveness of the cleaning process.

[0080] A machine learning algorithm or tool (308) uses a first subset of training data (306) to identify correlations between identified features (e.g., one or more cleaning process parameters) that influence the corresponding known cleaning outcomes. In other words, the machine learning tool uses the first subset of training data to train a cleaning outcome classifier (310). The first subset of training data is also used to tune (312) the machine learning model to improve or maximize its performance. The machine learning tool uses a second subset of training data to evaluate or score how well the cleaning outcome classifier can predict cleaning outcomes. The result of the training is a trained cleaning outcome classifier (316).

[0081] The cleaning result classifier (316) can be used to perform tasks as described in this paper. Figure 9Evaluation of one or more new cleaning processes shown and described.

[0082] Figures 6A to 6C This is a graph illustrating exemplary results obtained from evaluations of different binary clean result classifiers and using different feature sets. Generally, the goal of a binary clean result classifier is to predict one of two possible responses—a "clean" result and a "contaminated" result. The confusion matrix is ​​a 2x2 table formed by counting the number of four results from the binary classifier. For the purposes of this specification, positive label = contaminated, and negative label = clean. An exemplary confusion matrix is ​​shown in Table 2.

[0083] Table 2

[0084]

[0085] Various metrics can be derived from the confusion matrix, and these metrics can be used to evaluate the accuracy of the binary clean outcome classifier. The error rate is calculated as the number of all incorrect predictions divided by the total number of data points. The best error rate is 0.0, and the worst error rate is 1.0. Accuracy is calculated as the number of all correct predictions divided by the total number of data points. The best accuracy is 1.0, and the worst accuracy is 0.0. Accuracy can also be calculated using the error rate of 1. Precision is calculated as the number of correct positive predictions divided by the total number of positive predictions. The best precision is 1.0, and the worst precision is 0.0. The Matthews correlation coefficient and F-score can also be calculated for each binary clean outcome classifier.

[0086] For example, Figure 6A A graph showing the true positive rate versus false positive rate of an exemplary cleanliness outcome classifier generated using a two-level (binary) logistic regression model with feature set A (see Table 1) is presented. Figure 6A The lower part shows the various statistics calculated for this model, including the number of true positives (TP), false positives (FP), false negatives (FN), and true negatives (TN).

[0087] The logistic regression model also generates a list of features and weights, which can be used to evaluate the importance of each feature to the outcome within the predictive model. Figure 6A The table on the right illustrates these. High positive values ​​indicate greater importance in predicting positive labels (contaminated test strips), while large negative values ​​indicate greater importance in predicting negative labels (clean test strips).

[0088] Generally speaking, Figure 6AThe two-stage logistic regression model achieved an accuracy of 0.812, meaning that the model accurately predicted "clean" or "soiled" 81.2% of the time. In this example, the most important feature was thermal conductivity, followed by wash time and rinse temperature (negative weights indicate they contribute more to the negative label = clean). The least important feature was wash temperature.

[0089] Figure 6B A graph showing the true positive rate versus false positive rate of an exemplary cleanliness outcome classifier generated using a two-level boosting decision tree model with feature set A (see Table 1) is presented. Figure 6B The lower part shows the various statistics calculated for this model, including the number of true positives (TP), false positives (FP), false negatives (FN), and true negatives (TN), accuracy, precision, recall, F1 score, and area under the curve. For comparison with... Figure 6A Using the same feature set as the model, the accuracy statistic of this model is 0.916. Therefore, based on Figure 6A and Figure 6B The computation of the model, when using feature set A in this example, appears that the two-level boosting decision tree model performs better than the two-level logistic regression model (accuracy = 0.812).

[0090] Figure 6C A graph showing the true positive rate versus false positive rate of an exemplary cleanliness outcome classifier generated using a two-level boosting decision tree model with feature set D (see Table 1) is presented. Figure 6B The lower part shows the various statistics calculated for this model, including the number of true positives (TP), false positives (FP), false negatives (FN), and true negatives (TN), accuracy, precision, recall, F1 score, and area under the curve. For comparison with... Figure 6A Using the same feature set as the model, the accuracy statistic of this model is 0.948. Therefore, based on Figure 6B and Figure 6C In the computation of the model, in this example, the two-level boosting decision tree model using feature set D performs better than the two-level logistic boosting decision tree model using feature set A (accuracy = 0.812).

[0091] exist Figure 6C The table on the right shows the importance of features. According to the model, detergent concentration was identified as the most important feature for predicting a "clean" result, followed by water hardness titration, thermal conductivity, washing time, washing temperature, rinsing temperature, rinsing aid concentration, rinsing time, detergent type, rinsing aid type, and food stains.

[0092] Figures 6A to 6CThe examples are given as examples of different machine learning models and different feature sets that can be used to generate clean result classifiers according to the techniques of the present invention. It should be understood that these examples are not intended to be limiting, and other combinations of other machine learning models and feature sets can be used, and this disclosure is not limited in this respect.

[0093] Can target Figures 6A to 6C Other statistical values ​​determined by the exemplary model include, but are not limited to:

[0094] -Accuracy = (Correctly predicted categories / Total test categories) * 100 = ((TP + TN) / (TP + TN + FP + FN)) * 100);

[0095] Precision = (True Positives / Total Predicted Positives) * 100 = (TP / (TP + FP)) * 100). This statistic is an indicator of the model's accuracy. It can be useful when the cost of false positives is high (e.g., clean test strips are labeled as contaminated).

[0096] - Recall = (True Positives / Total Actual Positives) * 100 = (TP / (TP + FN)) * 100. This statistic indicates how many actual positives the model captured by labeling them as positive. This statistic can be useful when the cost of false negatives is high (e.g., contaminated test strips are predicted to be clean).

[0097] - F1 score = 2 * (precision * recall / (precision + recall)) - used to find the balance between precision and recall; useful when there is a non-uniform class distribution (e.g., a large number of true negatives).

[0098] -AUC = Area under the curve. This statistic indicates the degree to which the model can distinguish between clean and contaminated categories.

[0099] - True Positive Rate (TPR) - The number of positive results classified as positive by the algorithm divided by the total number of positive results.

[0100] - False positive rate (FPR) - The number of negatives classified as positive by the algorithm divided by the total number of negatives; FPR = FP / (TN + FP).

[0101] These and other statistics can also be calculated for other machine learning models, and it should be understood that this disclosure is not limited in this respect.

[0102] Figure 7This is a diagram illustrating a summary of exemplary classification model results from several two-level classification model tools according to this disclosure. The classification models include two-level logistic regression models, two-level boosting decision tree models, two-level neural network models, two-level Bayesian point machine models, and two-level support vector machine (SVM) models. Exemplary results for each of these models are given for each of feature sets A, B, C, and D (see the list of feature sets in Table 1 above). In this example, the two-level boosting decision tree model gives the most accurate prediction for each feature set in the feature set.

[0103] Figure 7 Additional features that can be included in the training data are also shown: the rack location of the validation sample. For example, in some types of dishwashers, a validation sample placed in a specific location on the dishwasher rack may be more indicative of cleaning effectiveness than a validation sample placed in other rack locations. Therefore, the rack location corresponding to each validation sample can also be included as one of the features of the training data, along with one or more cleaning process parameters and known results (e.g., “clean” or “contaminated” or numerical scores).

[0104] For example, a validation sample placed in the rear left corner of a door-type commercial dishwasher may be a better indicator of cleaning effectiveness than one placed in other rack positions. Figure 7 The indicator is "Rack Location 1". When rack location is considered, the accuracy of the two-level logistic regression model increases for all feature sets. In this particular example, when rack location is considered, the accuracy of the two-level boosting decision tree model decreases for all feature sets. This is likely because, in this particular example, the decision tree model overfits the data due to the small number of data points. It should be understood that this disclosure is not limited in this respect, and the examples shown are for the purpose of illustrating an exemplary process of selecting from different available machine learning models.

[0105] In other examples, machine learning models that employ regression to generate quantified values ​​or numerical scores of cleaning results can also be used. Figure 8This is a graph illustrating an overview of exemplary regression model results from several regression modeling tools according to this disclosure. The regression models include linear regression models, boosting decision tree regression models, neural network regression models, and Bayesian linear regression models. Exemplary results for each of these models are given for each of feature sets A, B, C, and D (see the list of feature sets in Table 1 above). In this example, the boosting decision tree regression model gives the most accurate predictions for feature sets C and D, taking into account all rack positions (0.891). Rack positions include four samples in three distinct locations across the rack: position 1 in the lower left corner of the rack, positions 5A and 5B in the center of the rack, and position 3 in the lower right corner of the rack. When only rack position 1 is considered, the accuracy of the boosting decision tree regression model increases to 0.926. This is likely due to the fact that rack position 1 is the most difficult to clean in this particular type of cleaning machine due to obstructions in front of the spray path or other obstructions or inconsistencies within the cleaning chamber.

[0106] Figure 7 and Figure 8 Many different machine learning models are illustrated, and different combinations of feature sets can be used to train cleaning outcome classifiers. Depending on the type of machine, the artifacts to be cleaned, and other factors, different machine learning models and / or different feature sets can generate optimal cleaning outcome predictions. Therefore, it should be understood that any machine learning model can be used in place of the machine learning model described herein, and this disclosure is not limited in this respect. Furthermore, it should be understood that different combinations of feature sets and / or additional or alternative features can replace the specific feature set described herein, and this disclosure is not limited in this respect.

[0107] Figure 9 This is a flowchart illustrating an exemplary process (350) according to the present disclosure, in which a computing device classifies the results of a new cleaning process performed by a cleaning machine using a trained cleaning result classifier. The computing device may include, for example... Figure 1 or Figure 4 An exemplary cleaning machine controller 200, and can control the process (350) based on the execution of instructions stored in the cleaning process control module 212 and the trained cleaning result classifier and executed by the processor 202.

[0108] At the start of a new cleaning process (352), the computing device uses the stored cleaning process parameters to control the execution of the new cleaning process (354). The stored cleaning process parameters may be stored in a storage device, such as a storage device forming part of the cleaning machine controller. Figure 4 Storage device 208 of the cleaning machine controller 200 shown.

[0109] The computing device monitors one or more cleaning process parameters (356) during the execution of the cleaning process. The one or more cleaning process parameters monitored during the cleaning process may include those from the machine itself or sensors associated with the cleaning machine (such as…). Figure 4 The parameters measured by the sensor 220 shown include washing temperature, rinsing temperature, washing time, rinsing time, and thermal conductivity.

[0110] One or more cleaning process parameters may also include product type parameters, such as detergent type and / or rinsing agent type, which are manually determined and stored in the cleaning machine controller. Detergent type and rinsing agent type may also be automatically determined, for example, by reading electronically readable codes (such as barcodes or QR codes) associated with the detergent and / or rinsing agent dispensed by the product dispensing system.

[0111] One or more cleaning process parameters may also include parameters determined by one or more manual testing procedures and stored in the cleaning machine controller, such as water hardness titration and / or alkalinity titration performed by a field service technician.

[0112] One or more cleaning process parameters may also include parameters indicating the presence of food contaminants in the wash water. For example, a food contaminant parameter could be a Boolean parameter indicating the presence of food contaminants in the cleaning solution (e.g., food contaminants "yes" or "no"). Food contaminants are commonly present in commercial establishments because they are typically found in storage tanks (e.g., such as...). Figure 1 The storage tank 110 shown contains at least a certain level of food contaminants. In another example, the food contaminant parameter can be assigned a numerical value representing the relative amount of food contaminants in the cleaning solution. For example, a turbidity measurement can be used as a representative of the level of food contaminants in the cleaning solution in the storage tank. For this purpose, sensor 220 may include a turbidity sensor or other sensors that measure a parameter indicating the amount of food contaminants present in the cleaning solution in the storage tank. In another example, if fresh water is used for each cleaning process instead of reusing the cleaning solution from the storage tank, the food contaminant parameter can be set to "No" or a numerical value indicating that there is no food contaminants in the cleaning solution.

[0113] Once the cleaning process is complete (358), the computing device stores cycle data corresponding to the cleaning process (360). The cycle data includes one or more cleaning process parameters that are monitored or otherwise correspond to the cleaning process during its execution. As described above, the cleaning process parameters may include one or more of the following: washing temperature, rinsing temperature, washing time, rinsing time, thermal conductivity, detergent type, rinsing aid type, water hardness titration, alkalinity titration, food stains, and / or any other parameters that may affect the effectiveness of the cleaning process.

[0114] The computing device classifies or scores cleaning results using a trained cleaning result classifier based on selected cleaning process parameters from one or more cleaning process parameters monitored during the execution of the cleaning process (362). The selected cleaning process parameters include a set of features used as input to the trained cleaning result classifier. The cleaning process parameters used to classify or score the results of a new cleaning process can be the same as those used to train the cleaning result classifier during the training phase.

[0115] When the trained cleaning result classifier classifies or rates the cleaning result as "clean" or assigns a score indicating a "clean" result (the "Yes" branch of 364), process (300) completes (368). When the trained cleaning result classifier classifies the cleaning result as "contaminated" or assigns a score indicating a "contaminated" cleaning result (e.g., a score less than a threshold) (the "No" branch of 364), the computing device adjusts the stored cleaning process parameters to ensure satisfactory cleaning results for subsequent cleaning processes performed by the cleaning machine (366). For example, the computing device may predict cleaning result classifications or scores for one or more hypothetical cleaning processes, each using a different set of adjusted cleaning process parameters. The computing device can then select an adjusted set of cleaning process parameters that results in a "clean" prediction for the cleaning result classification or score to be used for one or more subsequent cleaning processes.

[0116] Figure 10 This is a flowchart illustrating an exemplary process (370) according to the present invention, in which a computing device uses a trained cleaning process classifier to predict the cleaning result of a current new cleaning process, and dynamically adjusts one or more cleaning process parameters during the execution of the current cleaning process to ensure a satisfactory cleaning result. The computing device may include, for example... Figure 1 or Figure 4 An exemplary cleaning machine controller 200 can control the process (370) based on the execution of instructions stored in the cleaning process control module 212 and the trained cleaning result classifier and executed by the processor 202.

[0117] At the start of a new cleaning process (372), the computing device uses the stored cleaning pass parameters to control the execution of the current new cleaning process (374). The stored cleaning process parameters may be stored in a storage device, such as a storage device forming part of the cleaning machine controller. Figure 4 Storage device 208 of the cleaning machine controller 200 shown.

[0118] The computing device monitors one or more cleaning process parameters (376) during the execution of the current new cleaning process. The one or more cleaning process parameters monitored during the current new cleaning process may include those described above. Figure 9 The parameters discussed include, for example, washing temperature, rinsing temperature, washing time, rinsing time, and thermal conductivity; and product type parameters, such as detergent type and / or rinsing agent type, which are manually determined and stored in the cleaning machine controller. Detergent type and rinsing agent type can also be automatically determined, for example, by reading electronically readable codes (such as barcodes or QR codes) associated with detergents and / or rinsing agents dispensed by the product dispensing system; parameters determined by one or more manual testing procedures and stored in the cleaning machine controller (such as water hardness titration and / or alkalinity titration performed by a field service technician); parameters indicating the presence of food contaminants in the wash water; and measurements of the presence of food contaminants in the water.

[0119] One or more cleaning process parameters may be measured once or multiple times during the execution of the cleaning process. For example, one or more cleaning process parameters may be measured continuously at a predetermined sampling rate during the execution of the cleaning process. Some cleaning process parameters may be measured at different times or at different rates, or at a single time point, or before or after the cleaning process.

[0120] During one or more times during the execution of the current new cleaning process, the computing device may classify or score the cleaning result using a trained cleaning result classifier based on one or more monitored cleaning process parameters associated with time (378). For example, at a predetermined time after the start of the cleaning process, the computing device may classify or score the cleaning result using a trained cleaning result classifier based on one or more cleaning process parameters monitored at or before the predetermined time (378). The predetermined time may be, for example, a predetermined number of seconds after the start of the cleaning process, such as 5 seconds, 10 seconds, 15 seconds, or other predetermined number of seconds after the start of the cleaning process. If the predicted result based on the cleaning process parameters associated with the predetermined time is “contaminated” or unsatisfactory (the “No” branch of 380), the computing device may dynamically adjust the cleaning process parameters to ensure a satisfactory cleaning result for the current new cleaning process (390). The computing device then controls the remainder of the current new cleaning process based on the adjusted cleaning process parameters (392).

[0121] As another example, the computing device may use a trained cleaning outcome classifier to classify or score the cleaning outcome based on one or more monitored cleaning process parameters measured during each of one or more sampling periods (378). For example, if the sampling period is 1 second, the computing device may predict the classification or score of the cleaning outcome associated with each 1-second sampling period. If the prediction for any one or more sampling periods is “contaminated” or unsatisfactory (the “No” branch of 380), the computing device may dynamically adjust the cleaning process parameters to ensure a satisfactory cleaning outcome for the current new cleaning process (390). Alternatively, the computing device may require a minimum number of sampling periods to have a corresponding “contaminated” cleaning outcome prediction before dynamically adjusting the cleaning process parameters for the current new cleaning process.

[0122] The adjusted cleaning process parameters can be determined by predicting the cleaning results of one or more different adjusted cleaning process parameters and selecting one set of adjusted cleaning process parameters from among the multiple sets of adjusted cleaning process parameters that result in a "clean" prediction for the current new cleaning process (390). The computing device then controls the remainder of the current new cleaning process based on the adjusted cleaning process parameters (392).

[0123] If the prediction result is “clean” or otherwise satisfactory at one or more predetermined times or for any one or more sampling periods (the “yes” branch of 380), the computing device continues to execute the current new cleaning process using the original cleaning process parameters (382).

[0124] Once the cleaning process is complete (384), the computing device stores cycle data (386) corresponding to the cleaning process. The cycle data includes one or more cleaning process parameters that are monitored or otherwise correspond to the cleaning process during its execution. As described above, the cleaning process parameters may include one or more of the following: washing temperature, rinsing temperature, washing time, rinsing time, thermal conductivity of the washing water, detergent type, rinsing aid type, water hardness of the washing water, alkalinity of the washing water, and / or measurements of the presence of food contaminants in the washing water and / or any other parameters that may affect the effectiveness of the cleaning process.

[0125] While the examples given herein are described with reference to automated cleaning machines (e.g., dishwashers or dishwashing machines) for food preparation / processing applications, it should be understood that the techniques described herein for classifying and / or scoring cleaning results can be applied to a variety of other applications. Such applications may include, for example, food and / or beverage processing equipment, laundry applications, agricultural applications, hospitality applications, and / or any other applications where cleaning, sterilization, or disinfection of items may be useful.

[0126] In one or more examples, the functionality described herein may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functionality may be stored as one or more instructions or code on or transmitted over a computer-readable medium and executed by a hardware-based processing unit. A computer-readable medium may comprise a computer-readable storage medium corresponding to a volatile medium such as a data storage medium or a communication medium that facilitates, for example, transferring a computer program from one place to another according to a communication protocol. In this way, a computer-readable medium may generally correspond to (1) a non-transitory tangible computer-readable storage medium or (2) a communication medium such as a signal or carrier wave. A data storage medium may be any available medium that can be accessed by one or more computers or one or more processors to retrieve instructions, code, and / or data structures for implementing the techniques described herein. Computer program products may include computer-readable media.

[0127] By way of example, and not limitation, such computer-readable storage media may include RAM, ROM, EEPROM, CD-ROM or other optical disc storage devices, magnetic disk storage devices or other magnetic storage devices, flash memory, or any other medium that can be used to store desired program code in the form of instructions or data structures and that can be accessed by a computer. Furthermore, any connection is appropriately referred to as a computer-readable medium. For example, if instructions are transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of medium. However, it should be understood that computer-readable storage media and data storage media do not include connections, carrier waves, signals, or other transient media, but rather refer to non-transient volatile storage media. As used, disks and optical discs include compact optical discs (CDs), laser discs, optical discs, digital versatile optical discs (DVDs), floppy disks, and Blu-ray discs, where disks typically reproduce data magnetically, while optical discs reproduce data optically using lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0128] The instructions can be executed by one or more processors, such as one or more digital signal processors (DSPs), general-purpose microprocessors, application-specific integrated circuits (ASICs), field-programmable arrays (FPGAs), or other equivalent integrated or discrete logic circuit systems. Therefore, the term "processor" as used can refer to any of the foregoing structures or any other structure suitable for implementing the described techniques. Additionally, in some examples, the described functionality can be housed within dedicated hardware and / or software modules. Furthermore, the techniques can be fully implemented in one or more circuit or logic elements.

[0129] The techniques disclosed herein can be implemented in a variety of devices or apparatuses, including wireless mobile phones, integrated circuits (ICs), or a set of ICs (e.g., chipsets). Various components, modules, or units are described in this disclosure to emphasize functional aspects of a device configured to perform the disclosed techniques, but they do not necessarily need to be implemented through different hardware units. Rather, as described above, the various units can be combined in a hardware unit or provided by a series of interoperable hardware units including one or more processors as described above, in conjunction with suitable software and / or firmware.

[0130] It should be recognized that, depending on the example, some actions or events in any of the methods described herein may be performed in a different order, may be added, combined, or omitted entirely (e.g., practicing the methods does not require all of the described actions or events). Furthermore, in some examples, actions or events may be performed simultaneously rather than sequentially, for example, through multithreading, interrupt handling, or multiple processors.

[0131] In some examples, computer-readable storage media may include non-transitory media. The term "non-transitory" may indicate that the storage medium is not embodied in a carrier wave or propagating signal. In some examples, non-transitory storage media may store data that may change over time (e.g., in RAM or cache memory).

[0132] Example

[0133] Example 1: An automated cleaning machine includes at least one processor; at least one storage device storing one or more predefined cleaning process parameters and a trained cleaning result classifier; the at least one storage device further includes instructions executable by the at least one processor to: control the cleaning machine to perform at least one cleaning process using the one or more predefined cleaning process parameters; monitor one or more cleaning process parameters during the execution of the cleaning process; classify or score the result of the cleaning process using the trained cleaning process classifier based on the one or more cleaning process parameters monitored during the execution of the cleaning process; and adjust one or more predefined cleaning process parameters in the predefined cleaning process parameters in response to the trained cleaning process classifier classifying the result of the cleaning process as contaminated, such that subsequent cleaning processes will be classified as clean by the trained cleaning result classifier.

[0134] Example 2: The automated cleaning machine according to Example 1, wherein the trained cleaning process classifier classifies the result of the cleaning process as either clean or contaminated.

[0135] Example 3: The automated cleaning machine according to Example 1, wherein the trained cleaning process classifier scores the result of the cleaning process by assigning numerical scores indicating the cleaning result.

[0136] Example 4: The automated cleaning machine according to Example 1, wherein the one or more cleaning cycle parameters include one or more of the following: washing temperature, rinsing temperature, washing time, rinsing time, thermal conductivity of washing water, detergent type, rinsing aid type, water hardness of washing water, alkalinity of washing water and / or measurement results of the presence of food dirt in washing water.

[0137] Example 5: The automated cleaning machine according to Example 4, wherein the measurement result of the presence of food dirt is a Boolean parameter that makes a first possible value of food dirt = true and a second possible value of food dirt = false.

[0138] Example 6: The automated cleaning machine according to Example 4, wherein the measurement result of the presence of food dirt includes the turbidity measurement result of the cleaning solution in the storage tank of the cleaning machine.

[0139] Example 7: The automated cleaning machine according to Example 1, wherein the trained cleaning result classifier is either a trained two-level classification machine learning model or a trained regression machine learning model.

[0140] Example 8: The automated cleaning machine according to Example 1, wherein the at least one storage device further includes instructions that can be executed by the at least one processor to control the cleaning machine to perform subsequent cleaning processes using one or more adjusted predefined cleaning process parameters.

[0141] Example 9: The automated cleaning machine according to Example 1, wherein the trained cleaning result classifier is trained using training data obtained from one or more designed experiments or field tests, in which one or more cleaning process verification test pieces are placed in the cleaning chamber of the cleaning machine and exposed to the cleaning process performed by the cleaning machine during the training phase.

[0142] Example 10: The automated cleaning machine according to Example 1, wherein the trained cleaning result classifier is trained based on one or more cleaning process parameters corresponding to each of a plurality of cleaning processes executed during the training phase and the known outputs corresponding to each of the plurality of cleaning processes executed during the training phase.

[0143] Example 11: A method comprising storing one or more predefined cleaning process parameters and a trained cleaning result classifier in a storage device of an automated cleaning machine; controlling the automated cleaning machine to perform at least one cleaning process using the one or more predefined cleaning process parameters via a controller of the automated cleaning machine; monitoring the one or more cleaning process parameters by the controller of the automated cleaning machine during the execution of the cleaning process; classifying or scoring the result of the cleaning process by the controller of the automated cleaning machine based on the one or more cleaning process parameters monitored during the execution of the cleaning process using the trained cleaning process classifier; and adjusting one or more predefined cleaning process parameters in the predefined cleaning process parameters by the controller of the automated cleaning machine in response to the trained cleaning process classifier classifying the result of the cleaning process as contaminated, such that subsequent cleaning processes will be classified as clean by the trained cleaning result classifier.

[0144] Example 12: According to the method of Example 11, wherein the trained cleaning process classifier classifies the result of the cleaning process as either clean or contaminated.

[0145] Example 13: According to the method of Example 11, wherein the trained cleaning process classifier scores the result of the cleaning process by assigning numerical scores indicating the cleaning result.

[0146] Example 14: According to the method described in Example 11, the one or more cleaning cycle parameters include one or more of the following: washing temperature, rinsing temperature, washing time, rinsing time, thermal conductivity of washing water, detergent type, rinsing aid type, water hardness of washing water, alkalinity of washing water and / or measurement results of the presence of food dirt in washing water.

[0147] Example 15: The method according to Example 14, wherein the measurement result of the presence of food contaminants is a Boolean parameter that makes a first possible value of food contaminants = true and a second possible value of food contaminants = false.

[0148] Example 16: The method according to Example 14, wherein the measurement result of the presence of food contaminants includes the turbidity measurement result of the cleaning solution in the storage tank of the cleaning machine.

[0149] Example 17: According to the method described in Example 11, the trained clean result classifier is either a trained two-level classification machine learning model or a trained regression machine learning model.

[0150] Example 18: The method according to Example 11 further includes using the one or more predefined cleaning process parameters to control the cleaning machine to perform at least one cleaning process.

[0151] Example 19: The method according to Example 11, wherein the trained cleaning outcome classifier is trained using training data obtained from one or more designed experiments or field tests, in which one or more cleaning process verification test pieces are placed in the cleaning chamber of a cleaning machine and exposed to the cleaning process performed by the cleaning machine during the training phase.

[0152] Example 20: The method according to Example 11, wherein the trained cleaning result classifier is trained based on one or more cleaning process parameters corresponding to each of the plurality of cleaning processes executed during the training phase and the known output corresponding to each of the plurality of cleaning processes executed during the training phase.

[0153] Example 21: An automated cleaning machine includes: at least one processor; at least one storage device storing one or more predefined cleaning process parameters and a trained cleaning result classifier; the at least one storage device further includes instructions executable by the at least one processor to: control the cleaning machine to perform at least one cleaning process using the one or more predefined cleaning process parameters; monitor one or more cleaning process parameters during the execution of the cleaning process; classify or score the result of the cleaning process using the trained cleaning process classifier based on the one or more cleaning process parameters monitored during the execution of the cleaning process; dynamically adjust one or more predefined cleaning process parameters in response to the trained cleaning process classifier classifying the result of the cleaning process as contaminated, such that the trained cleaning result classifier classifies the cleaning process as clean; and control the cleaning machine to perform the remaining portion of the cleaning process using the dynamically adjusted one or more predefined cleaning process parameters.

[0154] Various embodiments have been described. These and other embodiments are within the scope of the appended claims.

Claims

1. An automated cleaning machine, comprising: At least one processor; At least one storage device stores one or more predefined cleaning process parameters and a trained cleaning process classifier, wherein the trained cleaning process classifier is a trained two-level classification machine learning model and is configured to classify the result of the cleaning process of the automated cleaning machine as clean or contaminated. The at least one storage device further includes instructions executable by the at least one processor to perform the following operations: The automated cleaning machine is controlled to perform the cleaning process using one or more predefined cleaning process parameters. One or more cleaning process parameters are monitored during the execution of the cleaning process; Based on one or more cleaning process parameters monitored during the execution of the cleaning process, the results of the cleaning process are classified using the trained cleaning process classifier; and In response to the trained cleaning process classifier classifying the result of the cleaning process as contamination: The trained cleaning process classifier is used to predict the cleaning outcome of one or more hypothetical cleaning processes, which use different sets of adjusted cleaning process parameters; and Select one group from the adjusted cleaning process parameters that leads to a cleanliness prediction; and The automated cleaning machine controls the execution of subsequent cleaning processes using one of the selected groups of adjusted cleaning process parameters.

2. The automated cleaning machine according to claim 1, wherein, The one or more cleaning process parameters include one or more of the following: washing temperature, rinsing temperature, washing time, rinsing time, thermal conductivity of washing water, detergent type, rinsing aid type, water hardness of washing water, alkalinity of washing water, and / or measurement results of the presence of food dirt in washing water.

3. The automated cleaning machine according to claim 2, wherein, The measurement of the presence of food contaminants is a Boolean parameter that sets the first possible value of food contaminants to true and the second possible value of food contaminants to false.

4. The automated cleaning machine according to claim 2, wherein, Measurements of the presence of food contaminants include measurements of the turbidity of the cleaning solution in the storage tank of the automated cleaning machine.

5. The automated cleaning machine according to claim 1, wherein, The trained cleaning process classifier is trained using training data obtained from one or more designed experiments or field tests in which one or more cleaning process validation specimens are placed in the cleaning chambers of one or more cleaning machines and exposed to cleaning processes performed by the one or more cleaning machines during the training phase.

6. The automated cleaning machine according to claim 1, wherein, The trained cleaning process classifier is trained based on one or more cleaning process parameters corresponding to each of the plurality of cleaning processes executed during the training phase, and the known outputs corresponding to each of the plurality of cleaning processes executed during the training phase.

7. A method for automating a cleaning machine, comprising: One or more predefined cleaning process parameters and a trained cleaning process classifier are stored in the storage device of the automated cleaning machine, wherein the trained cleaning process classifier is a trained two-level classification machine learning model and is configured to classify the result of the cleaning process of the automated cleaning machine as clean or contaminated. The automated cleaning machine is controlled by its controller to perform at least one cleaning process using one or more predefined cleaning process parameters. During the execution of the cleaning process, one or more cleaning process parameters are monitored by the controller of the automated cleaning machine; The controller of the automated cleaning machine classifies the results of the cleaning process using the trained cleaning process classifier based on one or more cleaning process parameters monitored during the execution of the cleaning process; and In response to the trained cleaning process classifier classifying the result of the cleaning process as contamination: The trained cleaning process classifier is used to predict the cleaning outcome of one or more hypothetical cleaning processes, which use different sets of adjusted cleaning process parameters; and Select one group from the adjusted cleaning process parameters that leads to a cleanliness prediction; and The automated cleaning machine uses one of the selected groups of adjusted cleaning process parameters to control the subsequent cleaning process.

8. The method according to claim 7, wherein, The one or more cleaning process parameters include one or more of the following: washing temperature, rinsing temperature, washing time, rinsing time, thermal conductivity of washing water, detergent type, rinsing aid type, water hardness of washing water, alkalinity of washing water, and / or measurement results of the presence of food dirt in washing water.

9. The method according to claim 8, wherein, The measurement of the presence of food contaminants is a Boolean parameter that sets the first possible value of food contaminants to true and the second possible value of food contaminants to false.

10. The method according to claim 8, wherein, Measurements of the presence of food contaminants include measurements of the turbidity of the cleaning solution in the storage tank of the automated cleaning machine.

11. The method according to claim 7, wherein, The trained cleaning process classifier is trained using training data obtained from one or more designed experiments or field tests in which one or more cleaning process validation specimens are placed in the cleaning chambers of one or more cleaning machines and exposed to cleaning processes performed by the one or more cleaning machines during the training phase.

12. The method according to claim 7, wherein, The trained cleaning process classifier is trained based on one or more cleaning process parameters corresponding to each of the plurality of cleaning processes executed during the training phase, and the known outputs corresponding to each of the plurality of cleaning processes executed during the training phase.

13. An automated cleaning machine, comprising: At least one processor; At least one storage device stores one or more predefined cleaning process parameters and a trained cleaning process classifier, wherein the trained cleaning process classifier is a trained two-level classification machine learning model and is configured to classify the result of the cleaning process of the automated cleaning machine as clean or contaminated. The at least one storage device further includes instructions executable by the at least one processor to perform the following operations: The automated cleaning machine is controlled to perform the cleaning process using one or more predefined cleaning process parameters. One or more cleaning process parameters are monitored during the execution of the cleaning process; Based on one or more cleaning process parameters monitored during the execution of the cleaning process, the results of the cleaning process are classified using the trained cleaning process classifier. In response to the trained cleaning process classifier classifying the result of the cleaning process as contamination: The trained cleaning process classifier is used to predict the cleaning outcome of one or more hypothetical cleaning processes, which use different sets of adjusted cleaning process parameters; and Select one group from the adjusted cleaning process parameters that leads to a cleanliness prediction; and The automated cleaning machine is controlled to perform the remaining portion of the cleaning process using one of the selected groups of adjusted cleaning process parameters.

Images