EP0928929B1 - Automatic cooking device using a neural network - Google Patents

Automatic cooking device using a neural network Download PDF

Info

Publication number
EP0928929B1
EP0928929B1 EP19990400041 EP99400041A EP0928929B1 EP 0928929 B1 EP0928929 B1 EP 0928929B1 EP 19990400041 EP19990400041 EP 19990400041 EP 99400041 A EP99400041 A EP 99400041A EP 0928929 B1 EP0928929 B1 EP 0928929B1
Authority
EP
European Patent Office
Prior art keywords
cooking
cavity
neurones
temperature
dish
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Lifetime
Application number
EP19990400041
Other languages
German (de)
French (fr)
Other versions
EP0928929A1 (en
Inventor
Jean-Paul Chevrier
Pascal Oudart
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Europeenne De Fabrication D'enceintes Mi Cie
Brandt Cooking SAC
Brandt Industries SAS
Original Assignee
Compagnie Europeenne pour lEquipement Menager SA
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Compagnie Europeenne pour lEquipement Menager SA filed Critical Compagnie Europeenne pour lEquipement Menager SA
Publication of EP0928929A1 publication Critical patent/EP0928929A1/en
Application granted granted Critical
Publication of EP0928929B1 publication Critical patent/EP0928929B1/en
Anticipated expiration legal-status Critical
Expired - Lifetime legal-status Critical Current

Links

Images

Classifications

    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24C—DOMESTIC STOVES OR RANGES ; DETAILS OF DOMESTIC STOVES OR RANGES, OF GENERAL APPLICATION
    • F24C7/00—Stoves or ranges heated by electric energy
    • F24C7/08—Arrangement or mounting of control or safety devices
    • F24C7/087—Arrangement or mounting of control or safety devices of electric circuits regulating heat

Definitions

  • the invention relates to the field of cooking devices automatic for oven.
  • the invention applies preferably to ovens traditional.
  • the purpose of automatic cooking devices is to simplify maximum life of the user while guaranteeing the best result cooking possible.
  • the object of the invention is to propose a device based on the use of a neural network, which does not require the establishment of a system of empirical rules that are otherwise laborious to design and unsuitable for the cooking operation.
  • the device according to the invention does not use preferably two physical measurements, the temperature prevailing in the cavity and moisture emitted by the food during cooking, which does not requires no complex weighing devices in the oven. It's a very automated device because the only information required of the user is the family of dishes, no other information is needed for a good cooking process, which requires little effort on the part of the user.
  • the dish family corresponds to the nature of the food, for example "chicken" or "pie”.
  • an automatic cooking device having an oven, at least one temperature sensor measuring the temperature in the oven cavity, at least one humidity sensor measuring moisture in the oven cavity, and a network of neurons, selection means to which the user provides a given family of dishes information, characterized in that that the device also comprises means to associate with the neural network a set of link weights between the neurons, the game being adapted to the dish family, means of which launch a cooking method adapted to the dish family, extraction means which extract a group of parameters from the measurements made by the temperature sensor and the sensor of moisture, in that the neural network estimates a cooking time remaining from the parameter group, and that the cooking means perform the remaining cooking time in an open loop.
  • FIG. 1 schematically shows a cooking device automatic according to the invention.
  • the arrows represent exchanges of data between the different parts of the device, the letters or groups of letters near the arrows symbolically represent the data transmitted.
  • This device comprises a cavity 10 of oven, at least one temperature sensor 11, at least one humidity sensor 12, means 13 cooking.
  • the plate 14 is introduced into the cavity 10 to be cooked.
  • the temperature sensor 11 makes it possible to measure the temperature T in the cavity 10
  • the humidity sensor makes it possible to measure the humidity H in the cavity 10.
  • the sensors 11 and 12 can be located in the cavity 10, but it is not obligatory.
  • the temperature sensor 11 is located in cavity 10, while the two moisture sensors are located in an air guide shown in Figure 2 and connecting the cavity 10 to external environment and allowing ventilation of the cavity 10.
  • the device can have other types of sensors, but these are sufficient.
  • the device also comprises means 13 for cooking, which usually comprise heating elements not shown on the Fig.
  • heating elements can be elements heated by the ground, that is to say by the underside of the cavity 10, or by the grill, that is to say by the upper face of the cavity 10, or heating elements arranged around a ventilation system of the cavity 10, or a combination of the above heating elements.
  • the 13 cooking means can heat the cavity 10 and cook the dish 14 via heating elements.
  • the attachments of the elements 11, 12 and 13 to the cavity 10 are represented by dashed lines.
  • the device also comprises means 20 for extracting parameters which develop from the measurements made by the temperature sensors 11 and 12 humidity a GP group of parameters which will be detailed later.
  • the device also comprises a network 30 of neurons which receives in input group GP of previous parameters and which outputs to cooking means 13 a cooking time TCR remaining
  • the device also includes selection means 40 by which the user 50 provides the device FP family information flat, that is to say that the user indicates the category of food to which the dish belongs intended for cooking.
  • the selection means 40 may for example consist of a keyboard or a set of buttons, each button corresponding to a family of dishes. Dish families can by example be "pie”, “chicken”, “soup”, or other families define according to the particular application envisaged.
  • the families of FP dishes are preferably chosen so that the dishes of the same family have relatively homogeneous properties in cooking.
  • the plate 14 is introduced into the cavity 10.
  • the FP family information of dishes is provided to the means 40 for selection by the user 50.
  • the user triggers the cooking operation, pressing example on an on / off button.
  • FP information is transmitted by means 40 for selecting the cooking means 13.
  • the means 13 of start a cooking mode adapted to the FP family of dishes, by example for a family of dishes "pie" heating will be done especially by the floor heating element.
  • the FP information is transmitted to the network 30 neurons by the selection means 40.
  • the neuron network 30 is then configured in a manner adapted to the family of dishes FP, which is preferably homogeneous.
  • the neural network 30 has several successive layers of neurons 34, three layers 31 to 33 for example in Figure 1.
  • the neurons 34 of a given layer, for example the layer 32 are linked to the neurons 34 of the neighboring layers, the layers 31 and 33, by links 35.
  • each of these links is associated with a weighting, i.e. coefficient, all of these weights constituting a set of weights.
  • To each family of dishes FP corresponds a game of weights.
  • a weighting game adapted to the FP family of dishes is associated with the neural network 30 which is therefore configured in a manner suitable for the FP family of dishes.
  • the FP information will also be provided to the means 20 for extracting parameters which will also be configured appropriately for the FP family of dishes.
  • the two sensors 11 and 12 respectively go take measurements of temperature and humidity. Preferably, measurements are made from the beginning of cooking. From these measurements, the extraction means 20 will extract after a certain time which depends preferably on the family of flat FP a group of parameters GP which will be detailed later.
  • the parameter group GP is injected into input, that is to say on the side of the layer 31, the network 30 of neurons.
  • the neuron network 30 estimates a cooking time TCR remaining from the GP parameter group.
  • the TCR time is transmitted to the cooking means 13 which perform in loop open the remaining cooking time TCR.
  • Open loop means that from the moment when the time TCR has been transmitted to the cooking means 13, these are disconnected from the network 30 of neurons, perform the cooking time TCR remaining without any change in mode or cooking time is provided by the network 30 of neurons until the end of cooking.
  • the various means mentioned above are functional representations, and that the device may comprise a microprocessor responsible for carrying out all or part of the operations previously described as well as coordinate them.
  • Other links can conventionally exist between the various means of the device, as for example that between the temperature sensor 11 and the means cooking 13 allowing the cooking means 13 to control the cooking; they are not shown in Figure 1.
  • the group of parameters consists of the initial temperature Ti of the cavity 10, the derivative dT of the temperature of the cavity 10 with respect to the time, the flow rate. of water emitted by the dish at a first moment t 1 and the quantity Qe of water emitted by the dish from the beginning of cooking to a second time t 2 .
  • the initial temperature of the cavity 10 is the temperature in the cavity 10 at the beginning of cooking.
  • the derivative dT is the slope of the temperature over a certain period, it reflects the thermal mass of the plate 14, the thermal mass being the product mass (or weight) by calorisfique capacity; as inside the same family of dishes FP, the heat capacity is assumed to evolve similarly regardless of the flat 14, the derivative dT strongly depend on the weight of the flat 14. For example, in the family of chicken dish, from one chicken to another, the evolution of the dT derivative will depend mainly on the weight of the chicken.
  • the parameter of the derivative dT advantageously replaces the missing weight information of the plate 14:
  • the flow rate De corresponds to the humidity prevailing in the cavity at time t 1 .
  • the quantity Qe is the quantity of water emitted from the beginning of cooking to a time t 2 ; Qe therefore corresponds to the integration over time of the humidity prevailing in the cavity.
  • the instants t 1 and t 2 preferably depend on the FP plate family, they are optimized by performing tests.
  • the plate family FP is therefore data transmitted to the extraction means for extracting the parameter group GP.
  • these instants t 1 and t 2 coincide, which facilitates the operation of extracting the parameters.
  • the choice of these parameters results from a quality optimization of the estimation of the neuron network with respect to the complexity of the GP parameter group and the structure of the neuron network.
  • FIGs 2A and 2B show schematically respectively a top view and a side view of an air guide of an embodiment preferential of an automatic cooking device according to the invention.
  • the automatic cooking device uses an already existing air guide 1 for the ventilation of the oven cavity during cooking.
  • Air Guide 1 is located above the upper face of the cavity not shown on the Fig. Guide 1 allows ventilation of the cavity by evacuating air loaded with moisture from the water emitted by the dish during cooking.
  • the air leaving the cavity passes through an unrepresented inlet of the guide 1 before to arrive in the guide 1.
  • the arrows in solid lines represent the air circulation ambient and the triangles connected by dotted lines represent the flow of air Cooking.
  • the device comprises a first sensor 3 of humidity placed in zone 2 of guide 1.
  • the first sensor 3 is subjected to thermal stresses much less important than if it were placed in the cavity, which allows the use of a moisture sensor that does not support only slight thermal stresses.
  • this first sensor 3 measures the humidity contained in a mixture containing cooking air certainly, but also ambient air whose humidity is different from that of the cavity.
  • the device then comprises a second humidity sensor 6 placed in the zone 5 measuring only the ambient humidity, that is to say the external environment.
  • Zones 2 and 5 are represented symbolically, the outline exact of said areas being more complex, and the boundary between the said areas not being brutal.
  • the device includes advantageously two temperature sensors 4 and 7 respectively placed in the vicinity of the humidity sensors 3 and 6, close enough so that the temperature differences between where is placed a moisture sensor and the place where is placed the annex sensor of associated temperature are insignificant.
  • Sensor information Annexes 4 and 7 then allow the device to modulate the response of humidity sensors 3 and 6 depending on the temperatures at which the Humidity sensors 3 and 6 are subject.
  • the temperature sensor measuring the temperature of the cavity is advantageously placed in the cavity, it will also serve as a probe of temperature control for cooking.
  • the oven preferably has a ventilation system to homogenize the temperature in the cavity during cooking.
  • the neural network of Figure 1 will now be detailed in a preferred embodiment.
  • the numbers are those of Figure 1.
  • the 30 network of neurons comprises neurons 34 distributed in several layers 31 to 33.
  • the network has three layers of neurons, the layer 31, the intermediate layer 32, and the exit layer 33.
  • neurons 34 are interconnected by links 35.
  • the links 35 have weights.
  • the neurons of the input layer 31 correspond parameters of the GP parameter group, so they are preferably number of four.
  • the output layer 33 provides a single value, the time TCR cooking remaining, so it is preferably composed of a single neuron.
  • Each neuron in the middle layer 32 receives from each of the four neurons of the input layer 31 a signal.
  • the signal issued by a neuron na from the input layer 31 to a neuron nb of the layer 32 intermediate is the value of the parameter assigned to the neuron na weighted by the weighting of the existing link between na neurons and nb.
  • the neuron of layer 33 of output receives signals from the neurons of layer 32 intermediate, signals from which it emits a signal representing time TCR cooking remaining estimated.
  • each family of dishes FP corresponds a set of weightings of the links between neurons .
  • These games of Weights are determined during learning phases on examples with the help of a cook expert. The different examples as well by variations of dish, while remaining in the same family, as initial temperature variations, for example.
  • Each weighting game is advantageously determined in the following manner. On a series examples belonging to the same family of dishes, the comparison between the cooking time given by the expert cook and the estimated cooking time by the neural network, we then obtain a error. The expert cook estimates the cooking time on the basis of his sight, his sense of smell, his experience, etc.
  • the neural network is striving to correct, to minimize errors by a statistical method seeking the correlations between the different examples.
  • the set of weights obtained when errors are minimized will be the game retained for the dish family corresponding, for the cooking then carried out by the user.
  • the network is said to be convergent.
  • Each family of dishes having a specific weighting set, on a common neural network structure one can thus define several "virtual" neural networks, that is to say differ only in their game of weights, each "virtual" neural network being adapted and therefore optimized for a particular FP family of dishes. This allows to obtain a Neural network structure both simple and convergent.
  • Each family of dishes may require a number of more examples or less, depending on whether the family is "simple" or "complex".
  • a simple family of dishes is a family whose different elements have a almost identical behavior
  • a complex family of dishes is a family whose different elements have a more disparate behavior. For as the neural network converges, the number of neurons in the intermediate layer is sufficiently large.
  • the network will tend to "to specialize", ie to associate each neuron of the layer intermediate to one or two examples of learning, and during a cooking performed by the user, the neural network will estimate a time of cooking remains wrong if the dish and cooking conditions do not match not exactly one of the examples of the learning phase. It takes contrary force the neural network to "generalize" the examples of the learning phase by choosing a ratio between the number of examples and the number of neurons in the intermediate layer that is sufficiently great.
  • the intermediate layer comprises six neurons, and the number of examples per family of dishes during the learning phase is of the order of twenty.
  • the device according to the invention is simple and effective. It also the advantage of being scalable. Indeed, the introduction of a new family of dishes only requires storage of a new set of weights obtained during an additional learning phase on a few examples of dishes belonging to the new family.
  • the device according to the invention can also combine the means previously described for families of relatively complex dishes and other more traditional means, for example using relationships directly calculating the remaining cooking time from the measurements made by temperature and humidity sensors, for families very simple dishes, that is to say of which all the elements have a uniform behavior.

Landscapes

  • Engineering & Computer Science (AREA)
  • Chemical & Material Sciences (AREA)
  • Combustion & Propulsion (AREA)
  • Mechanical Engineering (AREA)
  • General Engineering & Computer Science (AREA)
  • Cookers (AREA)
  • Electric Ovens (AREA)

Description

L'invention concerne le domaine des dispositifs de cuisson automatique pour four. L'invention s'applique de préférence aux fours traditionnels. Le but des dispositifs de cuisson automatique est de simplifier au maximum la vie de l'utilisateur tout en lui garantissant le meilleur résultat de cuisson possible.The invention relates to the field of cooking devices automatic for oven. The invention applies preferably to ovens traditional. The purpose of automatic cooking devices is to simplify maximum life of the user while guaranteeing the best result cooking possible.

Pour cela, il existe différents dispositifs dans l'état de l'art. Ces dispositifs effectuent en général plusieurs mesures physiques dans la cavité du four, puis ils appliquent une série de règles empiriques, par exemple à base de logique floue, afin de réguler le processus de cuisson sans intervention de l'utilisateur. Ces dispositifs ont l'inconvénient de nécessiter souvent la mesure automatique du poids de l'aliment qui est placé dans la cavité du four, ce qui est coûteux et difficile à mettre en oeuvre. Ils ont encore l'inconvénient soit d'être peu efficaces, l'établissement des règles empiriques étant malaisé, soit de demander l'intervention de l'utilisateur pour fournir un certain nombre de paramètres, ce qui aboutit à une cuisson insuffisamment automatisée et requérant une intervention trop importante de l'utilisateur.For this, there are different devices in the state of the art. These devices typically perform several physical measurements in the cavity of the furnace, then they apply a series of rules of thumb, for example to fuzzy logic base, to regulate the cooking process without user intervention. These devices have the disadvantage of requiring often the automatic measurement of the weight of the food that is placed in the oven cavity, which is expensive and difficult to implement. They still have the disadvantage of being inefficient, the establishment of rules of thumb being difficult, to request the intervention of the user to provide a number of parameters, resulting in insufficient cooking automated and requiring too much user intervention.

Il existe également, dans l'état de l'art, certains dispositifs utilisant un réseau de neurones pour évaluer certains paramètres difficiles d'accès, comme la température au centre de l'aliment pendant la cuisson. L'estimation par un réseau de neurones de ce type de paramètres reste insuffisante pour obtenir un bon résultat de cuisson, sur une gamme étendue d'aliments. comme le dispositif selon EP-A-529644 qui divulgue les caractéristiques de préambule de la revendication 1.There are also, in the state of the art, certain devices using a neural network to evaluate some hard-to-access parameters, like the temperature at the center of the food during cooking. The estimate by a neural network of this type of parameter remains insufficient for get a good cooking result on a wide range of foods. as the device according to EP-A-529644 which discloses the preamble features of claim 1.

Le but de l'invention est de proposer un dispositif basé sur l'utilisation d'un réseau de neurones, ce qui ne nécessite pas l'établissement d'un système de règles empiriques par ailleurs laborieuses à concevoir et inadaptées à l'opération de cuisson. Le dispositif selon l'invention n'utilise de préférence que deux mesures physiques, la température régnant dans la cavité et l'humidité émise par l'aliment au cours de la cuisson, ce qui ne nécessite pas de dispositifs complexes de pesée dans le four. C'est un dispositif très automatisé car la seule information requise de l'utilisateur est la famille de plats, aucune autre information n'est nécessaire pour un bon déroulement de la cuisson, ce qui exige peu d'effort de la part de l'utilisateur. La famille de plat correspond à la nature de l'aliment, par exemple « poulet » ou « tarte ».The object of the invention is to propose a device based on the use of a neural network, which does not require the establishment of a system of empirical rules that are otherwise laborious to design and unsuitable for the cooking operation. The device according to the invention does not use preferably two physical measurements, the temperature prevailing in the cavity and moisture emitted by the food during cooking, which does not requires no complex weighing devices in the oven. It's a very automated device because the only information required of the user is the family of dishes, no other information is needed for a good cooking process, which requires little effort on the part of the user. The dish family corresponds to the nature of the food, for example "chicken" or "pie".

Selon l'invention, il est prévu un dispositif de cuisson automatique comportant un four, au moins un capteur de température mesurant la température dans la cavité du four, au moins un capteur d'humidité mesurant l'humidité dans la cavité du four, et un réseau de neurones, des moyens de sélection auxquels l'utilisateur fournit une information de famille de plats donnee, caractérisé en ce que le dispositif comporte également des moyens pour associer au réseau de neurones un jeu de pondérations des liaisons entre les neurones, le jeu étant adapté à la famille de plat, des moyens de cuisson qui lancent un mode de cuisson adapté à la famille de plat, des moyens d'extraction qui extraient un groupe de paramètres à partir des mesures effectuées par le capteur de température et par le capteur d'humidité, en ce que le réseau de neurones estime un temps de cuisson restant à partir du groupe de paramètres, et en ce que les moyens de cuisson effectuent en boucle ouverte le temps de cuisson restant.According to the invention, there is provided an automatic cooking device having an oven, at least one temperature sensor measuring the temperature in the oven cavity, at least one humidity sensor measuring moisture in the oven cavity, and a network of neurons, selection means to which the user provides a given family of dishes information, characterized in that that the device also comprises means to associate with the neural network a set of link weights between the neurons, the game being adapted to the dish family, means of which launch a cooking method adapted to the dish family, extraction means which extract a group of parameters from the measurements made by the temperature sensor and the sensor of moisture, in that the neural network estimates a cooking time remaining from the parameter group, and that the cooking means perform the remaining cooking time in an open loop.

L'invention sera mieux comprise et d'autres caractéristiques et avantages apparaítront à l'aide de la description ci-après et des dessins joints, donnés à titre d'exemples non limitatifs, où:

  • la figure 1 représente schématiquement un dispositif de cuisson automatique selon l'invention.
  • la figure 2A représente schématiquement une vue de dessus d'un guide d'air d'un dispositif de cuisson automatique selon l'invention.
  • la figure 2B représente schématiquement une vue de côté d'un guide d'air d'un dispositif de cuisson automatique selon l'invention.
The invention will be better understood and other features and advantages will become apparent from the following description and the accompanying drawings, given by way of non-limiting examples, in which:
  • Figure 1 shows schematically an automatic cooking device according to the invention.
  • Figure 2A schematically shows a top view of an air guide of an automatic cooking device according to the invention.
  • Figure 2B schematically shows a side view of an air guide of an automatic cooking device according to the invention.

La figure 1 représente schématiquement un dispositif de cuisson automatique selon l'invention. Les flèches représentent des échanges de données entre les différentes parties du dispositif, les lettres ou groupes de lettres près des flèches représentent symboliquement les données transmises. Ce dispositif comporte une cavité 10 de four, au moins un capteur 11 de température, au moins un capteur 12 d'humidité, des moyens 13 de cuisson. Le plat 14 est introduit dans la cavité 10 pour être cuit. Pendant l'opération de cuisson, le capteur 11 de température permet de mesurer la température T dans la cavité 10, le capteur d'humidité permet de mesurer l'humidité H dans la cavité 10. Les capteurs 11 et 12 peuvent être situés dans la cavité 10, mais ce n'est pas obligatoire. Dans un mode de réalisation préférentiel de l'invention comportant un capteur 11 de température et deux capteurs 12 d'humidité , le capteur 11 de température, est situé dans la cavité 10, tandis que les deux capteurs d'humidité sont situés dans un guide d'air représenté sur la figure 2 et reliant la cavité 10 au milieu extérieur et permettant la ventilation de la cavité 10. Le dispositif peut comporter d'autres types de capteurs, mais ceux-ci sont suffisants. Le dispositif comporte également des moyens 13 de cuisson, lesquels comportent habituellement des éléments chauffants non représentés sur la figure. Par exemple, les éléments chauffants peuvent être des éléments chauffants par le sol, c'est-à-dire par la face inférieure de la cavité 10, ou par le grill, c'est-à-dire par la face supérieure de la cavité 10, ou bien des éléments chauffants disposés autour d'un système de ventilation de la cavité 10, ou encore une combinaison des éléments chauffants précédents. Les moyens 13 de cuisson permettent de chauffer la cavité 10 et de cuire le plat 14 par l'intermédiaire d'éléments chauffants. Les rattachements des éléments 11, 12 et 13 à la cavité 10 sont représentés par des traits pointillés. Le dispositif comporte également des moyens 20 d'extraction de paramètres qui élaborent à partir des mesures effectuées par les capteurs 11 de température et 12 d'humidité un groupe GP de paramètres qui seront détaillés plus loin. Le dispositif comporte également un réseau 30 de neurones qui reçoit en entrée le groupe GP de paramètres précédent et qui fournit en sortie aux moyens 13 de cuisson un temps TCR de cuisson restant Le dispositif comporte également des moyens 40 de sélection par lesquels l'utilisateur 50 fournit au dispositif une information FP de famille de plats, c'est-à-dire que l'utilisateur indique la catégorie d'aliment à laquelle appartient le plat 14 destiné à la cuisson. Les moyens 40 de sélection peuvent par exemple consister en un clavier ou en un ensemble de boutons, chaque bouton correspondant à une famille de plats. Les familles de plat peuvent par exemple être du type « tarte », « poulet », « soupe », ou d'autres familles à définir selon l'application particulière envisagée. Les familles de plats FP sont de préférence choisies pour que les plats d'une même famille aient des propriétés relativement homogènes en matière de cuisson.Figure 1 schematically shows a cooking device automatic according to the invention. The arrows represent exchanges of data between the different parts of the device, the letters or groups of letters near the arrows symbolically represent the data transmitted. This device comprises a cavity 10 of oven, at least one temperature sensor 11, at least one humidity sensor 12, means 13 cooking. The plate 14 is introduced into the cavity 10 to be cooked. During the cooking operation, the temperature sensor 11 makes it possible to measure the temperature T in the cavity 10, the humidity sensor makes it possible to measure the humidity H in the cavity 10. The sensors 11 and 12 can be located in the cavity 10, but it is not obligatory. In a mode of preferred embodiment of the invention comprising a sensor 11 of temperature and two humidity sensors 12, the temperature sensor 11, is located in cavity 10, while the two moisture sensors are located in an air guide shown in Figure 2 and connecting the cavity 10 to external environment and allowing ventilation of the cavity 10. The device can have other types of sensors, but these are sufficient. The device also comprises means 13 for cooking, which usually comprise heating elements not shown on the Fig. For example, heating elements can be elements heated by the ground, that is to say by the underside of the cavity 10, or by the grill, that is to say by the upper face of the cavity 10, or heating elements arranged around a ventilation system of the cavity 10, or a combination of the above heating elements. The 13 cooking means can heat the cavity 10 and cook the dish 14 via heating elements. The attachments of the elements 11, 12 and 13 to the cavity 10 are represented by dashed lines. The device also comprises means 20 for extracting parameters which develop from the measurements made by the temperature sensors 11 and 12 humidity a GP group of parameters which will be detailed later. The device also comprises a network 30 of neurons which receives in input group GP of previous parameters and which outputs to cooking means 13 a cooking time TCR remaining The device also includes selection means 40 by which the user 50 provides the device FP family information flat, that is to say that the user indicates the category of food to which the dish belongs intended for cooking. The selection means 40 may for example consist of a keyboard or a set of buttons, each button corresponding to a family of dishes. Dish families can by example be "pie", "chicken", "soup", or other families define according to the particular application envisaged. The families of FP dishes are preferably chosen so that the dishes of the same family have relatively homogeneous properties in cooking.

Le plat 14 est introduit dans la cavité 10. L'information FP de famille de plats est fournie aux moyens 40 de sélection par l'utilisateur 50. De préférence, l'utilisateur déclenche l'opération de cuisson, en appuyant par exemple sur un bouton marche/arrêt. L'information FP est transmise par les moyens 40 de sélection aux moyens 13 de cuisson. Les moyens 13 de cuisson lancent un mode de cuisson adapté à la famille de plats FP, par exemple pour une famille de plats « tarte » le chauffage s'effectuera surtout par l'élément chauffant de sol. L'information FP est transmise au réseau 30 de neurones par les moyens de sélection 40. Le réseau 30 de neurones est alors configuré d'une manière adaptée à la famille de plats FP, laquelle est de préférence homogène. Un avantage supplémentaire du dispositif est de comporter « virtuellement », sur une structure commune de réseau, plusieurs réseaux de neurones simples adaptés chacun à une famille de plats FP, au lieu d'un seul réseau global qui serait plus complexe et/ou moins efficace. « virtuellement» signifie ici ne différant que par leur jeu de pondérations. Le réseau 30 de neurones comporte plusieurs couches successives de neurones 34, trois couches 31 à 33 par exemple sur la figure 1. Les neurones 34 d'une couche donnée, par exemple la couche 32, sont liés aux neurones 34 des couches voisines, les couches 31 et 33, par des liaisons 35. A chacune de ces liaisons 35 est associée une pondération, c'est-à-dire un coefficient, l'ensemble des ces pondérations constituant un jeu de pondérations. A chaque famille de plats FP correspond un jeu de pondérations. La constitution précise du réseau 30 de neurones sera détaillée plus loin. Un jeu de pondérations adapté à la famille de plats FP est associé au réseau 30 de neurones lequel est donc configuré d'une manière adaptée à la famille de plats FP. De manière préférentielle, l'information FP sera également fournie aux moyens 20 d'extraction de paramètres lesquels seront alors aussi configurés de manière adaptée à la famille de plats FP.The plate 14 is introduced into the cavity 10. The FP family information of dishes is provided to the means 40 for selection by the user 50. Preferably, the user triggers the cooking operation, pressing example on an on / off button. FP information is transmitted by means 40 for selecting the cooking means 13. The means 13 of start a cooking mode adapted to the FP family of dishes, by example for a family of dishes "pie" heating will be done especially by the floor heating element. The FP information is transmitted to the network 30 neurons by the selection means 40. The neuron network 30 is then configured in a manner adapted to the family of dishes FP, which is preferably homogeneous. An additional advantage of the device is "virtually", on a common network structure, several networks of single neurons each adapted to a family of dishes FP, to instead of a single global network that would be more complex and / or less efficient. "Virtually" here means differing only in their weighting. The neural network 30 has several successive layers of neurons 34, three layers 31 to 33 for example in Figure 1. The neurons 34 of a given layer, for example the layer 32, are linked to the neurons 34 of the neighboring layers, the layers 31 and 33, by links 35. each of these links is associated with a weighting, i.e. coefficient, all of these weights constituting a set of weights. To each family of dishes FP corresponds a game of weights. The precise constitution of the neural network 30 will be detailed below. A weighting game adapted to the FP family of dishes is associated with the neural network 30 which is therefore configured in a manner suitable for the FP family of dishes. Preferably, the FP information will also be provided to the means 20 for extracting parameters which will also be configured appropriately for the FP family of dishes.

Pendant la cuisson, les deux capteurs 11 et 12 vont respectivement effectuer des mesures de température et d'humidité. De préférence, les mesures sont effectuées depuis le début de la cuisson. A partir de ces mesures, les moyens 20 d'extraction vont extraire au bout d'un certain temps qui dépend de préférence de la famille de plat FP un groupe de paramètres GP qui sera détaillé plus loin. Le groupe de paramètres GP est injecté en entrée, c'est-à-dire du côté de la couche 31, du réseau 30 de neurones. Le réseau 30 de neurones estime un temps TCR de cuisson restant à partir du groupe de paramètres GP. En sortie du réseau 30 de neurones, c'est-à-dire du côté de la couche 33, est fourni un temps TCR de cuisson restant. Le temps TCR est transmis aux moyens 13 de cuisson qui effectuent en boucle ouverte le temps TCR de cuisson restant. « en boucle ouverte » signifie qu'à partir du moment où le temps TCR a été transmis aux moyens 13 de cuisson, ces derniers sont déconnectés du réseau 30 de neurones, effectuent le temps TCR de cuisson restant sans qu'aucune modification de mode ou de durée de cuisson ne soit apportée par le réseau 30 de neurones jusqu'à la fin de la cuisson. Il est bien entendu que les différents moyens précédents sont des représentations fonctionnelles, et que le dispositif peut comporter un microprocesseur chargé de réaliser tout ou partie des opérations précédemment décrites ainsi que de les coordonner. D'autres liaisons peuvent classiquement exister entre les différents moyens du dispositif, comme par exemple celle entre le capteur de température 11 et les moyens de cuisson 13 permettant aux moyens de cuisson 13 de contrôler la cuisson ; elles ne sont pas représentées sur la figure 1.During cooking, the two sensors 11 and 12 respectively go take measurements of temperature and humidity. Preferably, measurements are made from the beginning of cooking. From these measurements, the extraction means 20 will extract after a certain time which depends preferably on the family of flat FP a group of parameters GP which will be detailed later. The parameter group GP is injected into input, that is to say on the side of the layer 31, the network 30 of neurons. The neuron network 30 estimates a cooking time TCR remaining from the GP parameter group. At the output of the neuron network 30, that is to say on the side of the layer 33, there is provided a cooking time TCR remaining. The TCR time is transmitted to the cooking means 13 which perform in loop open the remaining cooking time TCR. "Open loop" means that from the moment when the time TCR has been transmitted to the cooking means 13, these are disconnected from the network 30 of neurons, perform the cooking time TCR remaining without any change in mode or cooking time is provided by the network 30 of neurons until the end of cooking. It is understood that the various means mentioned above are functional representations, and that the device may comprise a microprocessor responsible for carrying out all or part of the operations previously described as well as coordinate them. Other links can conventionally exist between the various means of the device, as for example that between the temperature sensor 11 and the means cooking 13 allowing the cooking means 13 to control the cooking; they are not shown in Figure 1.

Dans un mode de réalisation préférentiel d'un dispositif selon l'invention, le groupe de paramètres est constitué de la température initiale Ti de la cavité 10, de la dérivée dT de la température de la cavité 10 par rapport au temps, du débit De d'eau émise par le plat à un premier instant t1 et de la quantité Qe d'eau émise par le plat depuis le début de la cuisson jusqu'à un deuxième instant t2. La température initiale de la cavité 10 est la température régnant dans la cavité 10 au début de la cuisson. La dérivée dT est la pente de la température sur une certaine période, elle traduit la masse thermique du plat 14, la masse thermique étant le produit masse (ou poids) par capacité calorisfique ; comme à l'intérieur d'une même famille de plats FP, la capacité calorifique est supposée évoluer de manière similaire quelque soit le plat 14, la dérivée dT dépendra alors fortement du poids du plat 14. Par exemple, au sein de la famille de plat « poulet », d'un poulet à l'autre, l'évolution de la dérivée dT dépendra surtout du poids du poulet. Le paramètre de la dérivée dT se substitue avantageusement à l'information manquante de poids du plat 14: Le débit De correspond à l'humidité régnant dans la cavité à l'instant t1. La quantité Qe est la quantité d'eau émise depuis le début de la cuisson jusqu'à un instant t2 ; Qe correspond donc à l'intégration au cours du temps de l'humidité régnant dans la cavité. Les instants t1 et t2 dépendent préférentiellement de la famille de plat FP, ils sont optimisés en réalisant des essais. La famille de plat FP est par conséquent une donnée transmise aux moyens 20 d'extraction pour l'extraction du groupe de paramètres GP. De préférence, ces instants t1 et t2 sont confondus, ce qui facilite l'opération d'extraction des paramètres. Le choix de ces paramètres résultent d'une optimisation qualité de l'estimation du réseau 30 de neurones par rapport à la complexité du groupe de paramètres GP et de la structure du réseau 30 de neurones.In a preferred embodiment of a device according to the invention, the group of parameters consists of the initial temperature Ti of the cavity 10, the derivative dT of the temperature of the cavity 10 with respect to the time, the flow rate. of water emitted by the dish at a first moment t 1 and the quantity Qe of water emitted by the dish from the beginning of cooking to a second time t 2 . The initial temperature of the cavity 10 is the temperature in the cavity 10 at the beginning of cooking. The derivative dT is the slope of the temperature over a certain period, it reflects the thermal mass of the plate 14, the thermal mass being the product mass (or weight) by calorisfique capacity; as inside the same family of dishes FP, the heat capacity is assumed to evolve similarly regardless of the flat 14, the derivative dT strongly depend on the weight of the flat 14. For example, in the family of chicken dish, from one chicken to another, the evolution of the dT derivative will depend mainly on the weight of the chicken. The parameter of the derivative dT advantageously replaces the missing weight information of the plate 14: The flow rate De corresponds to the humidity prevailing in the cavity at time t 1 . The quantity Qe is the quantity of water emitted from the beginning of cooking to a time t 2 ; Qe therefore corresponds to the integration over time of the humidity prevailing in the cavity. The instants t 1 and t 2 preferably depend on the FP plate family, they are optimized by performing tests. The plate family FP is therefore data transmitted to the extraction means for extracting the parameter group GP. Preferably, these instants t 1 and t 2 coincide, which facilitates the operation of extracting the parameters. The choice of these parameters results from a quality optimization of the estimation of the neuron network with respect to the complexity of the GP parameter group and the structure of the neuron network.

Les figures 2A et 2B représentent schématiquement respectivement une vue de dessus et une vue de côté d'un guide d'air d'une réalisation préférentielle d'un dispositif de cuisson automatique selon l'invention. Le dispositif de cuisson automatique utilise un guide d'air 1 déjà existant pour la ventilation de l'air de la cavité du four pendant la cuisson. Le guide 1 d'air est situé au-dessus de la face supérieure de la cavité non représentée sur la figure. Le guide 1 permet la ventilation de la cavité en évacuant de l'air chargé d'humidité provenant de l'eau émise par le plat pendant la cuisson. L'air quittant la cavité passe par une entrée non représentée du guide 1 avant d'arriver dans le guide 1. A l'air de cuisson provenant de la cavité, se joint dans le guide 1 de l'air ambiant provenant du milieu extérieur non représenté sur la figure. Les flèches en traits pleins représentent la circulation de l'air ambiant et les triangles reliés par pointillés représentent la circulation de l'air de cuisson. Sur le chemin de la circulation d'air, il y a une zone 2 dans laquelle circule un mélange d'air ambiant et d'air de cuisson. Il y a aussi une zone 5 dans laquelle ne circule que de l'air ambiant. Dans une réalisation préférentielle selon l'invention, le dispositif comporte un premier capteur 3 d'humidité placé dans la zone 2 du guide 1. Le premier capteur 3 est soumis à des contraintes thermiques beaucoup moins importantes que s'il était placé dans la cavité, ce qui autorise l'emploi d'un capteur d'humidité ne supportant que de faibles contraintes thermiques. Cependant, ce premier capteur 3 mesure l'humidité contenue dans un mélange contenant de l'air de cuisson certes, mais également de l'air ambiant dont l'humidité est différente de celle de la cavité. Afin de corriger cette influence de l'humidité du milieu extérieur, le dispositif comporte alors un deuxième capteur 6 d'humidité placé dans la zone 5 ne mesurant que l'humidité ambiante, c'est-à-dire du milieu extérieur. La connaissance des informations données respectivement par les deux capteurs 3 et 6 permet au dispositif de remonter à la valeur de l'humidité régnant dans la cavité, c'est-à-dire au débit d'eau émise par le plat pendant la cuisson. Les zones 2 et 5 sont représentées symboliquement, le contour exact des dites zones étant plus complexe, et la frontière entre les dites zones n'étant pas brutale.Figures 2A and 2B show schematically respectively a top view and a side view of an air guide of an embodiment preferential of an automatic cooking device according to the invention. The automatic cooking device uses an already existing air guide 1 for the ventilation of the oven cavity during cooking. Air Guide 1 is located above the upper face of the cavity not shown on the Fig. Guide 1 allows ventilation of the cavity by evacuating air loaded with moisture from the water emitted by the dish during cooking. The air leaving the cavity passes through an unrepresented inlet of the guide 1 before to arrive in the guide 1. To the cooking air coming from the cavity, joins in the guide 1 of the ambient air from the outside medium not shown on the face. The arrows in solid lines represent the air circulation ambient and the triangles connected by dotted lines represent the flow of air Cooking. On the way to the air circulation, there is an area 2 in which circulates a mixture of ambient air and cooking air. There is also a zone 5 in which circulates only ambient air. In one embodiment According to the invention, the device comprises a first sensor 3 of humidity placed in zone 2 of guide 1. The first sensor 3 is subjected to thermal stresses much less important than if it were placed in the cavity, which allows the use of a moisture sensor that does not support only slight thermal stresses. However, this first sensor 3 measures the humidity contained in a mixture containing cooking air certainly, but also ambient air whose humidity is different from that of the cavity. In order to correct this influence of the humidity of the external environment, the device then comprises a second humidity sensor 6 placed in the zone 5 measuring only the ambient humidity, that is to say the external environment. The knowledge of the information given respectively by the two sensors 3 and 6 allows the device to go back to the value of humidity in the cavity, that is to say the flow of water emitted by the dish during the cooking. Zones 2 and 5 are represented symbolically, the outline exact of said areas being more complex, and the boundary between the said areas not being brutal.

Les capteurs d'humidité pouvant avoir une réponse dépendant de la température à laquelle ils fonctionnent, le dispositif comporte avantageusement deux capteurs annexes 4 et 7 de température respectivement placés au voisinage des capteurs 3 et 6 d'humidité, suffisamment près pour que les différences de température entre l'endroit où est placé un capteur d'humidité et l'endroit où est placé le capteur annexe de température associé soient insignifiantes. Les informations des capteurs annexes 4 et 7 permettent alors au dispositif de moduler la réponse des capteurs d'humidité 3 et 6 en fonction des températures auxquelles les capteurs d'humidité 3 et 6 sont soumis.Humidity sensors that may have a response depending on the temperature at which they operate, the device includes advantageously two temperature sensors 4 and 7 respectively placed in the vicinity of the humidity sensors 3 and 6, close enough so that the temperature differences between where is placed a moisture sensor and the place where is placed the annex sensor of associated temperature are insignificant. Sensor information Annexes 4 and 7 then allow the device to modulate the response of humidity sensors 3 and 6 depending on the temperatures at which the Humidity sensors 3 and 6 are subject.

Le capteur de température mesurant la température de la cavité est avantageusement placé dans la cavité, il servira également de sonde de régulation de température pour la cuisson. Le four comporte de préférence un système de ventilation permettant d'homogénéiser la température dans la cavité pendant la cuisson.The temperature sensor measuring the temperature of the cavity is advantageously placed in the cavity, it will also serve as a probe of temperature control for cooking. The oven preferably has a ventilation system to homogenize the temperature in the cavity during cooking.

Le réseau de neurones de la figure 1 va maintenant être détaillé dans un mode de réalisation préférentiel. Les numéros sont ceux de la figure 1. Le réseau 30 de neurones comporte des neurones 34 réparties en plusieurs couches 31 à 33. Le réseau comporte trois couches de neurones, la couche 31 d'entrée, la couche 32 intermédiaire, et la couche de sortie 33. Les neurones 34 sont reliés entre eux par des liaisons 35. Il y a une liaison entre chaque neurone de la couche 31 d'entrée et chaque neurone de la couche 32 intermédiaire, ainsi qu'entre chaque neurone de la couche 32 intermédiaire et chaque neurone de la couche 33 de sortie. Les liaisons 35 ont des pondérations. Les neurones de la couche 31 d'entrée correspondent aux paramètres du groupe de paramètres GP, ils sont donc de préférence au nombre de quatre. La couche 33 de sortie fournit une seule valeur, le temps TCR de cuisson restant, elle est donc de préférence constituée d'un seul neurone. Chaque neurone de la couche 32 intermédiaire reçoit de chacun des quatre neurones de la couche 31 d'entrée un signal. Le signal émis par un neurone na de la couche 31 d'entrée vers un neurone nb de la couche 32 intermédiaire correspond à la valeur du paramètre affecté au neurone na pondérée par la pondération de la liaison existant entre les neurones na et nb. A son tour et dans les mêmes conditions, le neurone de la couche 33 de sortie reçoit des signaux provenant des neurones de la couche 32 intermédiaire, signaux à partir duquel il émet un signal représentant le temps TCR de cuisson restant estimé.The neural network of Figure 1 will now be detailed in a preferred embodiment. The numbers are those of Figure 1. The 30 network of neurons comprises neurons 34 distributed in several layers 31 to 33. The network has three layers of neurons, the layer 31, the intermediate layer 32, and the exit layer 33. neurons 34 are interconnected by links 35. There is a connection between each neuron of the input layer 31 and each neuron of the layer 32 as well as between each neuron in layer 32 intermediate and each neuron of the output layer 33. The links 35 have weights. The neurons of the input layer 31 correspond parameters of the GP parameter group, so they are preferably number of four. The output layer 33 provides a single value, the time TCR cooking remaining, so it is preferably composed of a single neuron. Each neuron in the middle layer 32 receives from each of the four neurons of the input layer 31 a signal. The signal issued by a neuron na from the input layer 31 to a neuron nb of the layer 32 intermediate is the value of the parameter assigned to the neuron na weighted by the weighting of the existing link between na neurons and nb. In turn, under the same conditions, the neuron of layer 33 of output receives signals from the neurons of layer 32 intermediate, signals from which it emits a signal representing time TCR cooking remaining estimated.

Comme expliqué plus haut, à chaque famille de plats FP correspond un jeu de pondérations des liaisons entre neurones.. Ces jeux de pondérations sont déterminés au cours de phases d'apprentissage sur des exemples avec l'aide d'un expert cuisinier. Les exemples différent aussi bien par des variations de plat, tout en restant dans la même famille, que des variations de température initiale par exemple. Chaque jeu de pondération est déterminé avantageusement de la façon suivante. Sur une série d'exemples appartenant à une même famille de plats, on effectue la comparaison entre le temps de cuisson donné par l'expert cuisinier et le temps de cuisson estimé par le réseau de neurones, on obtient alors une erreur. L'expert cuisinier estime le temps de cuisson sur la base de sa vue, de son odorat, de son expérience, etc. Au fil des exemples, par « rétropropagation du gradient de l'erreur », le réseau de neurones s'efforce de corriger, de minimiser ses erreurs par une méthode statistique recherchant les corrélations entre les différents exemples. Le jeu des pondérations obtenu lorsque les erreurs sont minimisées sera le jeu retenu pour la famille de plat correspondante, pour les cuissons réalisées ensuite par l'utilisateur. Lorsque les erreurs sont suffisamment minimisées, le réseau est dit « convergent ». Chaque famille de plats ayant un jeu de pondérations spécifique, sur une structure de réseau de neurones commune on peut ainsi définir plusieurs réseaux de neurones « virtuels », c'est-à-dire ne différant que par leur jeu de pondérations, chaque réseau de neurones « virtuel » étant adapté et donc optimisé pour une famille de plats FP particulière. Ceci permet d'obtenir une structure de réseau de neurones à la fois simple et convergente.As explained above, to each family of dishes FP corresponds a set of weightings of the links between neurons .. These games of Weights are determined during learning phases on examples with the help of a cook expert. The different examples as well by variations of dish, while remaining in the same family, as initial temperature variations, for example. Each weighting game is advantageously determined in the following manner. On a series examples belonging to the same family of dishes, the comparison between the cooking time given by the expert cook and the estimated cooking time by the neural network, we then obtain a error. The expert cook estimates the cooking time on the basis of his sight, his sense of smell, his experience, etc. As examples, by "Backpropagation of the gradient of error", the neural network is striving to correct, to minimize errors by a statistical method seeking the correlations between the different examples. The set of weights obtained when errors are minimized will be the game retained for the dish family corresponding, for the cooking then carried out by the user. When the errors are sufficiently minimized, the network is said to be convergent. Each family of dishes having a specific weighting set, on a common neural network structure one can thus define several "virtual" neural networks, that is to say differ only in their game of weights, each "virtual" neural network being adapted and therefore optimized for a particular FP family of dishes. This allows to obtain a Neural network structure both simple and convergent.

Chaque famille de plats peut nécessiter un nombre d'exemples plus ou moins élevé, selon que la famille est « simple » ou « complexe ». Une famille de plats simple est une famille dont les différents éléments ont un comportement quasi-identique, et une famille de plats complexe est une famille dont les différents éléments ont un comportement plus disparate. Pour que le réseau de neurones converge, il faut que le nombre de neurones de la couche intermédiaire soit suffisamment grand. Mais si ce nombre devient de l'ordre de grandeur du nombre d'exemples par famille de plats utilisés pendant les phases d'apprentissage, le réseau aura tendance à « se spécialiser », c'est-à-dire à associer chaque neurone de la couche intermédiaire à un ou deux exemples d'apprentissage, et lors d'une cuisson effectuée par l'utilisateur, le réseau de neurones estimera un temps de cuisson restant erroné si le plat et les conditions de cuisson ne correspondent pas exactement à l'un des exemples de la phase d'apprentissage. Il faut au contraire obliger le réseau de neurones à « généraliser » les exemples de la phase d'apprentissage en choisissant un rapport entre le nombre d'exemples et le nombre de neurones de la couche intermédiaire qui soit suffisamment grand. De préférence, la couche intermédiaire comporte six neurones, et le nombre d'exemples par famille de plats pendant la phase d'apprentissage est de l'ordre de vingt.Each family of dishes may require a number of more examples or less, depending on whether the family is "simple" or "complex". A simple family of dishes is a family whose different elements have a almost identical behavior, and a complex family of dishes is a family whose different elements have a more disparate behavior. For as the neural network converges, the number of neurons in the intermediate layer is sufficiently large. But if that number becomes the order of magnitude of the number of examples per family of dishes used During the learning phases, the network will tend to "to specialize", ie to associate each neuron of the layer intermediate to one or two examples of learning, and during a cooking performed by the user, the neural network will estimate a time of cooking remains wrong if the dish and cooking conditions do not match not exactly one of the examples of the learning phase. It takes contrary force the neural network to "generalize" the examples of the learning phase by choosing a ratio between the number of examples and the number of neurons in the intermediate layer that is sufficiently great. Preferably, the intermediate layer comprises six neurons, and the number of examples per family of dishes during the learning phase is of the order of twenty.

Le dispositif selon l'invention est simple et efficace. Il a également l'avantage d'être évolutif. En effet, l'introduction d'une nouvelle famille de plats ne nécessite que le stockage d'un nouveau jeu de pondérations obtenu au cours d'une phase d'apprentissage supplémentaire sur quelques exemples de plats appartenant à la nouvelle famille.The device according to the invention is simple and effective. It also the advantage of being scalable. Indeed, the introduction of a new family of dishes only requires storage of a new set of weights obtained during an additional learning phase on a few examples of dishes belonging to the new family.

Le dispositif selon l'invention peut également combiner les moyens précédemment décrits pour des familles de plats relativement complexes et d'autres moyens plus traditionnels, utilisant par exemple des relations calculant directement le temps de cuisson restant à partir des mesures effectuées par les capteurs de température et d'humidité, pour des familles de plats très simples, c'est-à-dire dont tous les éléments ont un comportement uniforme.The device according to the invention can also combine the means previously described for families of relatively complex dishes and other more traditional means, for example using relationships directly calculating the remaining cooking time from the measurements made by temperature and humidity sensors, for families very simple dishes, that is to say of which all the elements have a uniform behavior.

Claims (10)

  1. Automatic cooking device comprising an oven, at least one temperature sensor (11) measuring the temperature (T) in the cavity (10) of the oven, at least one humidity sensor (12) measuring the humidity (H) in the cavity (10) of the oven, and a network (30) of neurones (34) and selection means (30) to which the user (50) supplies given information on the class of dish (FP), means for associating with the network (30) of neurones (34) a set of weightings for the links (35) between the neurones (34), the set being adapted to the class (FP) of dish, cooking means (13) which initiate a cooking mode adapted to the class (FP) of dish, extraction means (20) which extract a group (GP) of parameters from the measurements made by the temperature sensor (11) and by the humidity sensor (12), in that the network (30) of neurones (34) estimates a remaining cooking time (TCR) from the group (GP) of parameters, and in that the cooking means (13) implement the remaining cooking time (TCR) in open loop.
  2. Device according to Claim 1, characterised in that the device has two absolute humidity sensors (3, 6) and an air guide (1) which has a first zone (2) where the air coming from the cavity and from an external environment passes and a second zone (5) where no air coming from the cavity passes but only air coming from the external environment, the first humidity sensor (3) being placed in the first zone (2) and the second humidity sensor (6) being placed in the second zone (5).
  3. Device according to Claim 2, characterised in that the device also comprises two subsidiary temperature sensors (4, 7), each subsidiary temperature sensor (4, 7) being placed in the vicinity of one of the humidity sensors (3, 6), and in that the device modulates in terms of temperature the response of the humidity sensors (3, 6) by means of the subsidiary temperature sensors (4, 7).
  4. Device according to any one of the preceding claims, characterised in that the group (GP) of parameters consists of the initial temperature (Ti) of the cavity (10), the drift (dT) of the temperature in the cavity (10) with respect to time, the output (De) of water emitted by the dish (14) at a first time (t1) and the quantity (Qe) of water emitted by the dish (14) from the start of the cooking up to a second time (t2).
  5. Device according to Claim 2, characterised in that the first time (t1) and the second time (t2) are merged.
  6. Device according to any one of Claims 4 to 5, characterised in that the first and second times (t1, t2) are adapted to the class (FP) of dish.
  7. Device according to any one of the preceding claims, characterised in that the network (30) of neurones (34) consists of three successive layers (31, 32, 33), which are: the input layer (31) consisting of four neurones, the intermediate layer (32), and the output layer (33) consisting of one neurone.
  8. Device according to Claim 7, characterised in that, each set of weightings of the links (35) being determined during learning with a predefined number of examples, the ratio between the number of examples and the number of neurones in the intermediate layer is sufficiently great to prevent the network (30) of neurones (34) from specialising in the examples.
  9. Device according to Claim 8, characterised in that the number of examples is around 20 and in that the intermediate layer comprises six neurones.
  10. Device according to any one of the preceding claims, characterised in that the oven is a traditional oven.
EP19990400041 1998-01-08 1999-01-08 Automatic cooking device using a neural network Expired - Lifetime EP0928929B1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
FR9800135 1998-01-08
FR9800135A FR2773390B1 (en) 1998-01-08 1998-01-08 AUTOMATIC COOKING DEVICE USING A NEURON NETWORK

Publications (2)

Publication Number Publication Date
EP0928929A1 EP0928929A1 (en) 1999-07-14
EP0928929B1 true EP0928929B1 (en) 2003-04-16

Family

ID=9521607

Family Applications (1)

Application Number Title Priority Date Filing Date
EP19990400041 Expired - Lifetime EP0928929B1 (en) 1998-01-08 1999-01-08 Automatic cooking device using a neural network

Country Status (4)

Country Link
EP (1) EP0928929B1 (en)
DE (1) DE69906826T2 (en)
ES (1) ES2195522T3 (en)
FR (1) FR2773390B1 (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE10336114A1 (en) 2003-08-06 2005-02-24 BSH Bosch und Siemens Hausgeräte GmbH Cooking device with a tanning sensor device
ITPN20050020A1 (en) 2005-04-05 2006-10-06 Electrolux Professional Spa "FREEZER PERFECTED WITH NEUTRAL NETWORK"
EP2188572B1 (en) 2007-08-24 2014-06-25 Arçelik Anonim Sirketi a method for operating a cooking oven
CN105444222B (en) * 2015-12-11 2017-11-14 美的集团股份有限公司 Cooking control method, system, Cloud Server and the micro-wave oven of micro-wave oven
CN107965803A (en) * 2017-11-16 2018-04-27 广东永衡良品科技有限公司 A kind of intelligent detection device and its control method for reminding anti-dry

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA2077018C (en) * 1991-08-30 1997-04-15 Kazunari Nishii Cooking appliance
JPH05172334A (en) * 1991-10-21 1993-07-09 Matsushita Electric Ind Co Ltd Cooking implement
JP2936853B2 (en) * 1991-12-20 1999-08-23 松下電器産業株式会社 kitchenware
US5893051A (en) * 1994-09-27 1999-04-06 Matsushita Electric Industrial Co., Ltd. Method of estimating temperature inside material to be cooked and cooking apparatus for effecting same

Also Published As

Publication number Publication date
EP0928929A1 (en) 1999-07-14
DE69906826D1 (en) 2003-05-22
FR2773390B1 (en) 2000-03-24
ES2195522T3 (en) 2003-12-01
DE69906826T2 (en) 2004-03-04
FR2773390A1 (en) 1999-07-09

Similar Documents

Publication Publication Date Title
CA2855773C (en) Cooking device
US11732225B2 (en) User interface for recipe creation
US5389764A (en) Automatic cooking appliance employing a neural network for cooking control
US8147888B2 (en) Method for the intelligent continuous filling of a cooking device and cooking device therefor
CN107255295B (en) Heating device, cooking system, arithmetic unit and cooking support method
US20230344959A1 (en) Infrared toaster
EP2898799B1 (en) Cooking system with assisted use
EP0928929B1 (en) Automatic cooking device using a neural network
JP6542802B2 (en) Method and apparatus for controlling the cooking process of food
FR2940899A1 (en) SEQUENTIAL COOKING COOKING APPARATUS, SYSTEM AND METHOD RELATED THERETO
Young Food photography: From snapshots to great shots
CN106388640A (en) Method and computer program for controlling a fryer, and fryer arranged for carrying out such method
CN108594673A (en) Menu generation method and kitchen appliance
Donis-González et al. Color vision system to assess English walnut (Juglans Regia) kernel pellicle color
EP0609762B1 (en) Method for controlling the cooking in a cooking apparatus having browning- and microwave generating devices
EP3957220A1 (en) Method for controlling cooking time and associated cooking appliance
Szocs et al. Forks over Spoons: The impact of cutlery on calorie estimates
WO2002095521A1 (en) Temperature compensation in an electronic circuit toaster
FR2807147A1 (en) PROCEDURE FOR OPERATING A COOKING OVEN AND CORRESPONDING COOKING OVEN
JP2855901B2 (en) kitchenware
JP2936839B2 (en) kitchenware
KR102567569B1 (en) A syrup for tanghulu bubble coating
CN112965542B (en) Intelligent cooking equipment control method, control equipment and intelligent terminal
Place Mark’s favorite Quadrotriticale
Billock Historic Chicago Bakeries

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

AK Designated contracting states

Kind code of ref document: A1

Designated state(s): DE DK ES FR GB IT NL SE

AX Request for extension of the european patent

Free format text: AL;LT;LV;MK;RO;SI

17P Request for examination filed

Effective date: 20000106

AKX Designation fees paid

Free format text: DE DK ES FR GB IT NL SE

GRAG Despatch of communication of intention to grant

Free format text: ORIGINAL CODE: EPIDOS AGRA

17Q First examination report despatched

Effective date: 20020422

GRAG Despatch of communication of intention to grant

Free format text: ORIGINAL CODE: EPIDOS AGRA

GRAH Despatch of communication of intention to grant a patent

Free format text: ORIGINAL CODE: EPIDOS IGRA

GRAH Despatch of communication of intention to grant a patent

Free format text: ORIGINAL CODE: EPIDOS IGRA

GRAA (expected) grant

Free format text: ORIGINAL CODE: 0009210

AK Designated contracting states

Designated state(s): DE DK ES FR GB IT NL SE

REG Reference to a national code

Ref country code: GB

Ref legal event code: FG4D

Free format text: NOT ENGLISH

REF Corresponds to:

Ref document number: 69906826

Country of ref document: DE

Date of ref document: 20030522

Kind code of ref document: P

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: SE

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20030716

Ref country code: DK

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20030716

GBT Gb: translation of ep patent filed (gb section 77(6)(a)/1977)

Effective date: 20030715

PLBE No opposition filed within time limit

Free format text: ORIGINAL CODE: 0009261

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: NO OPPOSITION FILED WITHIN TIME LIMIT

26N No opposition filed

Effective date: 20040119

REG Reference to a national code

Ref country code: FR

Ref legal event code: TP

Ref country code: FR

Ref legal event code: CD

Ref country code: FR

Ref legal event code: CA

NLS Nl: assignments of ep-patents

Owner name: COMPAGNIE EUROPEENNE DE FABRICATION D'ENCEINTES MI

Effective date: 20071018

Owner name: BRANDT INDUSTRIES

Effective date: 20071018

NLT1 Nl: modifications of names registered in virtue of documents presented to the patent office pursuant to art. 16 a, paragraph 1

Owner name: BRANDT COOKING

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: GB

Payment date: 20110222

Year of fee payment: 13

REG Reference to a national code

Ref country code: FR

Ref legal event code: CD

Owner name: FAGORBRANDT SAS, FR

Effective date: 20110826

REG Reference to a national code

Ref country code: ES

Ref legal event code: PC2A

Owner name: FAGORBRANDT SAS

Effective date: 20111021

REG Reference to a national code

Ref country code: NL

Ref legal event code: TD

Effective date: 20111221

REG Reference to a national code

Ref country code: DE

Ref legal event code: R082

Ref document number: 69906826

Country of ref document: DE

Representative=s name: PRINZ & PARTNER PATENTANWAELTE RECHTSANWAELTE, DE

Effective date: 20120322

Ref country code: DE

Ref legal event code: R082

Ref document number: 69906826

Country of ref document: DE

Representative=s name: PRINZ & PARTNER MBB PATENTANWAELTE RECHTSANWAE, DE

Effective date: 20120322

Ref country code: DE

Ref legal event code: R081

Ref document number: 69906826

Country of ref document: DE

Owner name: GROUPE BRANDT, FR

Free format text: FORMER OWNER: BRANDT INDUSTRIES, RUEIL MALMAISON, FR

Effective date: 20120322

Ref country code: DE

Ref legal event code: R081

Ref document number: 69906826

Country of ref document: DE

Owner name: FAGORBRANDT SAS, FR

Free format text: FORMER OWNER: BRANDT INDUSTRIES, RUEIL MALMAISON, FR

Effective date: 20120322

GBPC Gb: european patent ceased through non-payment of renewal fee

Effective date: 20120108

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: GB

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20120108

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: NL

Payment date: 20130121

Year of fee payment: 15

REG Reference to a national code

Ref country code: NL

Ref legal event code: V1

Effective date: 20140801

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: NL

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20140801

REG Reference to a national code

Ref country code: FR

Ref legal event code: PLFP

Year of fee payment: 18

REG Reference to a national code

Ref country code: ES

Ref legal event code: PC2A

Owner name: GROUPE BRANDT

Effective date: 20160307

REG Reference to a national code

Ref country code: DE

Ref legal event code: R082

Ref document number: 69906826

Country of ref document: DE

Representative=s name: PRINZ & PARTNER MBB PATENTANWAELTE RECHTSANWAE, DE

Ref country code: DE

Ref legal event code: R081

Ref document number: 69906826

Country of ref document: DE

Owner name: GROUPE BRANDT, FR

Free format text: FORMER OWNER: FAGORBRANDT SAS, RUEIL-MALMAISON, FR

REG Reference to a national code

Ref country code: FR

Ref legal event code: TP

Owner name: GROUPE BRANDT, FR

Effective date: 20160420

REG Reference to a national code

Ref country code: FR

Ref legal event code: GC

Effective date: 20160426

REG Reference to a national code

Ref country code: FR

Ref legal event code: PLFP

Year of fee payment: 19

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: DE

Payment date: 20170210

Year of fee payment: 19

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: IT

Payment date: 20170123

Year of fee payment: 19

Ref country code: ES

Payment date: 20170103

Year of fee payment: 19

REG Reference to a national code

Ref country code: FR

Ref legal event code: PLFP

Year of fee payment: 20

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: FR

Payment date: 20180129

Year of fee payment: 20

REG Reference to a national code

Ref country code: DE

Ref legal event code: R119

Ref document number: 69906826

Country of ref document: DE

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: DE

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20180801

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: IT

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20180108

REG Reference to a national code

Ref country code: ES

Ref legal event code: FD2A

Effective date: 20190730

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: ES

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20180109